A Revised Hemodynamic Theory of Age-Related Macular Degeneration

Abstract

Age-related macular degeneration (AMD) afflicts one out of every 40 individuals worldwide, causing irreversible central blindness in millions. The transformation of various tissue layers within the macula in the retina has led to competing conceptual models of the molecular pathways, cell types, and tissues responsible for the onset and progression of AMD. A model that has persisted for over 6 decades is the hemodynamic, or vascular theory of AMD progression, which states that vascular dysfunction of the choroid underlies AMD pathogenesis. Here, we re-evaluate this hypothesis in light of recent advances on molecular, anatomic, and hemodynamic changes underlying choroidal dysfunction in AMD. We propose an updated, detailed model of hemodynamic dysfunction as a mechanism of AMD development and progression.

Age-related macular degeneration (AMD) (see Glossary) is the leading cause of blindness in industrialized nations, affecting an estimated 2.5% of the world’s population [1], and the main risk factors include age and smoking. AMD is progressive, with the first clinical manifestation being the development of drusen or subretinal drusenoid deposits (SDD), depending on their location relative to the layers of the eye. These extracellular deposits of lipoproteins and inflammatory constituents in isolation, are considered subclinical. Over time, the deposits expand and coalesce, at which stage the disease is termed ‘early’ or ‘intermediate’. The stage depends on the size, number and location of deposits in the eye and the presence or absence of pigmentary changes, indicative of RPE disease [2]. In natural history studies, drusen reabsorption is closely succeeded by progression to ‘advanced’, blinding stages of AMD [3]. Most typically, advanced or ‘late’ AMD manifests as geographic atrophy (GA), a neovascular, or ‘wet’ form. GA is considered the ‘default’ end stage of AMD pathogenesis, typified by progressive deterioration of essentially all retinal layers, as well as the choroid and the retinal pigmented epithelium (RPE) (Figure 1, Box 1). GA causes irreversible blindness in over 1 million Americans and no approved therapies exist yet to prevent, halt or reverse its effects.

Figure 1. The Retina-Choroid Interface in AMD: Schematic Diagram of the Choroid and Retina in the Eye.

Light entering from the right is focused on the outer retinal photoreceptors (PR). PR are supported by the retinal pigmented epithelium (RPE) and the choroidal vasculature. The innermost layer of the choroid is the choriocapillaris (CC), a thin, dense, highly anastomotic microvascular structure. Between the RPE and the CC is Bruch’s membrane (BM), a specialized multi-layered extracellular matrix that divides the ‘inside’ and ‘outside’ of the eye. The first clinical presentation of AMD is the presence of drusen, extracellular debris that most often develop basal to the RPE, in between CC lumens, and subretinal drusenoid deposits (SDD) that develop between the RPE and photoreceptor layer, and cause outer retinal degeneration. AMD progresses to either geographic atrophy (GA), the gradual atrophy of the choroid, RPE and photoreceptors, and neovascular AMD, aberrant angiogenesis of the choroid through ruptures in BM and the RPE to the PR. Leakage of blood constituents into the retina causes disorganization and swelling of the retina, frequently incompatible with vision.

Box 1. The Retinal Pigmented Epithelium.

The RPE performs multiple functions including recycling components of the visual cycle, phagocytosis and lysosomal degradation of expired photoreceptor outer segments, light absorption, maintenance of the outer blood-retinal barrier by tight intracellular junctions, maintenance of the photoreceptor and choroidal layers by secretion of trophic factors such as VEGF A, and innate immune cell activation. Currently, and for the past dozen years, the RPE has been considered the most crucial cell layer to understand AMD. Drusen, the predominant clinical characteristic of early AMD, are derived from RPE and interrupt the RPE monolayer. Drusen constituents are implicated in AMD pathogenesis such as complement factors, amyloid β, and double-stranded RNA, inducing inflammatory and degenerative effects in RPE [107]. Early onset hereditary macular degenerations such as Stargardt disease are caused by genetic mutations in genes such as ABCA4 leading to RPE atrophy [108]. In addition to atrophy emanating from RPE-dependent signals, overexpression of VEGFA by the RPE is sufficient to induce aberrant subclinical choroidal neovascularization (CNV) [109], which may represent an early preclinical stage of neovascular AMD.

Thus, the choroid and photoreceptor layers have often been viewed as bystanders to the pathological process, passively following cues from the degenerating RPE. The importance of this view is manifested by recent efforts to develop cell-based therapies for advanced AMD. Endeavors thus far have exclusively utilized differentiated RPE cells derived from human embryonic stem cells [110, 111]. However, AMD is not in all cases an RPE cell-based disease, but rather, a disease where the hemodynamic microenvironment plays a decisive role, so replacing degenerating RPE with “healthy” “RPE” may be a fruitless effort, like sowing seeds in barren ground.

In neovascular AMD, the other form of advanced AMD, immature blood vessels, most commonly emanating from the choroid, but occasionally from the retinal vasculature, invade the outer retina. Exudation from these aberrant vessels causes anatomic disruption of the normally highly organized retinal architecture, leading to vision loss. Incidence of neovascular AMD is approximately twice that of GA [4, 5]. Clinical management of neovascular AMD was transformed by the approval and broad adoption of therapeutics that target the vascular endothelial growth factor-A (VEGFA) signaling pathway, but significant challenges remain for treatment and maintaining vision gains in these patients (Box 2).

Box 2. Anti-VEGFA Therapies for Neovascular AMD.

These therapies include the monoclonal antibody (Fab antibody portion) ranibizumab (Lucentis®, Genentech), its related full-length humanized monoclonal antibody bevacizumab (Avastin®, Genentech), and the VEGF-trap (recombinant fusion protein of VEGFR1 and VEGFR2) aflibercept (Eylea®, Regeneron), and pegatinib (Macugen®, Bausch and Lomb), a VEGFA binding aptamer. Although intraocular anti-VEGFA therapies have had clinical success, many patients do not attain significant visual improvement [112, 113]. For example, in neovascular AMD, over half of anti-VEGF recipients do not achieve 20/40 (driving) vision [114]. The long term prognosis for these individuals is sobering, where despite treatment, nearly 25% (representing millions of people) [1] have 20/200 vision (legal blindness) or worse [115], a majority of eyes develop untreatable central retinal atrophy [115, 116], and most short-term gains in visual acuity are lost within 4 or 5 years [116, 117]. Although a causal relationship between prolonged anti-VEGF therapy treatment and vision loss in humans has not been established, adverse effects of VEGFA neutralization on multiple retinal cell types including retinal neurons have been observed in animal models [118–122], and in patients treated with anti-VEGFA drugs for several years [115, 123, 124]. Thus, delineating the underlying causes of advanced AMD represents a major global health need.

A third, recently described type of atrophy in AMD consists of outer retinal atrophy associated with SDD deposition [6]. However, the relationship between the choroid and this type of AMD is unclear, and therefore restrict our discussion to GA and neovascular AMD.

Here, we describe the anatomic and hemodynamic properties of the choroid and, in light of recent advances in the field, revisit the hemodynamic theory of AMD, which posits that blood flow disturbances are the root cause of AMD. Re-evaluating this theory, we propose that hemodynamic changes in the choroid represent a key determinant of local manifestations of AMD pathogenesis.

The Retina in Age-Related Macular Degeneration

Deterioration of macular photoreceptors in AMD is devastating for vision and quality of life. Juxtaposed to the photoreceptor layer is the RPE, whose apical extensions intercalate with the dense layer of photoreceptor outer segments, and whose basal layer resides on Bruch’s membrane, separating the outer retina from the choroid (Figure 1).

With the focus on the RPE as the principal cell layer in determining the fate of the retina in AMD, the hemodynamic theory of AMD pathogenesis has received far less attention from cell and molecular biologists. However, emerging data collectively suggest that local choroidal hemodynamics may play a crucial role in the manifestation of AMD pathologies.

Friedman’s Hemodynamic Theory - Support and Need for Re-Evaluation

The concept that aberrant choroidal blood flow is a pathogenic stimulus for AMD dates back to at least 1905, when Possek proposed atherosclerotic changes to the ocular blood supply as a cause of what is now termed neovascular AMD [7]. The modern hemodynamic theory of AMD proposed by Friedman in 1997 [8], and updated over the next decade [9–11], hypothesizes that AMD arises due to alterations in choroidal blood flow resulting from increased stiffening of Bruch’s membrane and sclera. Stiffening of these structures is thought to be due to the accumulation of lipoproteins in a manner similar to lipid deposition in arterial walls during atherosclerosis [8]. Such stiffening increases choroidal vascular resistance, which results in either an increase in choriocapillaris hydrostatic pressure, or a decrease in choroidal perfusion, depending on the relative resistances of the ophthalmic artery (which supplies the choroid) and the cerebral artery (Figure 2).

Figure 2. Friedman’s Hemodynamic Theory of AMD.

In 1998 Friedman formally postulated that increased stiffness of the sclera, (supportive fibrous structure that lies posterior to the choroid), and BM, promotes AMD pathogenesis. By this model, the progression towards ‘wet’ or ‘dry’ AMD depends on whether the ophthalmic artery (OA), which provides blood the retinal and choroidal circulations, or the middle cerebral artery (MCA) provides greater vascular resistance. In the case where the OA provides less resistance due to intimal constriction of the the MCA, results in elevated blood pressure in the choroidal vasculature. Because of increased scleral stiffness to the posterior, and intraocular pressure to the anterior, elevated pressure within the choroidal vasculature results in hydrostatic pressure. Friedman postulated that this elevated hydrostatic pressure prevents the clearance of debris from the RPE layer to the choroid, and that this was the cause of neovascular AMD. If, on the other hand, the OA provides more resistance due to intimal constriction of the MCA, the choroid is hypoperfused. The choroid thins or collapses between intraocular pressure and a rigid sclera. Friedman postulated that choroidal hypoperfusion caused hypoxia and was responsible for geographic atrophy.

The concept that choroidal hemodynamics contribute to AMD progression is supported by recent clinical and experimental evidence. Altered choroidal hemodynamic parameters, such as reduced choroidal blood flow [12–16] and focal hypoperfusion [17] have been observed in human AMD. Analyses of postmortem human donor eyes have found that areas of decreased choriocapillaris density extends beyond the margin of RPE loss [18–21], which implies that choriocapillaris hemodynamic changes precede manifestation of advanced AMD. Studies in eyes of human AMD patients using optical coherence tomography angiography (OCTA) have also found that perfusion deficits in choriocapillaris extend beyond the margin of RPE atrophy [22, 23]. Protein expression of endothelial nitric oxide synthase, a major blood flow regulator, is significantly reduced in AMD-stricken choroid relative to age-matched healthy eyes [24]. Collectively, these studies support that choroidal hemodynamics alterations are associated with, and may be an early cause of AMD.

However, Friedman’s theory does not account for the finding that neither elevated intraocular pressure, nor Bruch’s membrane stiffening, while associated with normal aging, are associated with AMD pathology [25]. Still, no studies have ruled out the possibility that local variation in stiffness might contribute to local flow disturbances. Friedman’s hypothesis that AMD manifests due to vascular resistance imbalances in larger arteries does not itself explain the exquisitely focal nature of drusen deposits, or AMD pathologies overall. Both GA and CNV have been observed concomitantly in the same eye in humans [26], which is difficult to reconcile with Friedman’s model of ophthalmic artery disturbances as being the discriminating factor between these pathologies. Similarly, choroidal hypoperfusion and atrophy have been reported in both neovascular and atrophic AMD [21], suggesting that at a macro-scale, hemodynamic changes might be insufficient to discriminate between AMD outcomes.

These limitations in Friedman’s theory could be reconciled by considering choroidal hemodynamics in a more discrete manner. Others have suggested AMD to be a hypoxic response to hypoperfusion of the choroidal watershed zones which localize to the macula [27, 28], and CNV frequently localizes to watershed zones [29]. Postmortem analysis of human donor eyes also suggests AMD pathologic features spatially correlate with hemodynamic properties of choriocapillaris on a much finer length scale. For example, reduced density of capillaries and accumulation of ghost capillaries within the choriocapillaris are associated with drusen formation and disease severity [30]. Moreover, AMD pathologic features such as deposition of complement-related proteins, C-reactive protein and advanced glycation end products demonstrate punctate distribution in the RPE and Bruch’s membrane that spatially correlates to the spaces between vessels of the choriocapillaris, even in unaffected eyes [31–34].

One explanation for these observations is that spatial distribution of biochemical factors and drusen implicated in AMD are selectively deposited due to local heterogeneity in hemodynamic parameters. Taking this concept to its logical conclusion, we hypothesize that local variation in hemodynamic parameters, such as wall shear stress and residence time within the choriocapillaris govern the exquisitely focal nature of AMD pathogenesis, which is a centrally defining, yet unexplained, feature of the disease correlating strongly with retinal function [35].

An important caveat to this hypothesis is that local hemodynamic variabilities that exist within choriocapillaris alone are not sufficient to cause AMD. Choroidal ischemia in hypertensive choroidopathy, where blood flow within the choroid is diminished or absent, and Alzheimer’s disease where choroidal thinning is observed [36], are distinct entities from AMD. Moreover, computational modeling predicts that within the healthy choriocapillaris, highly heterogeneous flow rates and pressures exist which can vary by orders of magnitude within tens of microns [37, 38]. In analogy to the effects of local hemodynamics of large arteries on atherosclerotic plaque development, we predict that local hemodynamics within choriocapillaris are critical determinants of the location, severity and progression of disease, but in isolation, are not sufficient to explain all manifestations of the complex disease that is AMD.

Description of Choroidal Hemodynamics

The choroid performs multiple functions necessary for vision, among which are nutrient transport, waste removal and immune cell trafficking [39]. Macular photoreceptors, whose oxygen demand is among the largest in the human body [40], derive the majority of oxygen via diffusion from the choroidal circulation. Transport between the choroid and RPE/retina occurs at the level of the choriocapillaris, a highly anastomotic and extremely dense vascular bed. In the submacular choriocapillaris, the vascular lumen accounts for 80% of the volume of the tissue. The choriocapillaris has be computationally modeled as a thin (7–10 µm) sheet of blood interrupted by intercapillary pillars [37, 38]. Blood is supplied and drained from the choriocapillaris by small and mid-sized arterioles and venules that comprise Sattler’s layer (Figure 3). Despite high blood velocities in the choroidal arterioles and venules, empirical measurements of blood flow within the choriocapillaris suggest an average blood velocity far slower, with significant variability observed within the plane of the layer [41]. This may be due to the unique anatomic arrangement of the arteriorole, capillary, and venule network within the posterior pole and macula of the eye. Here, the arterioles and venules join the choriocapillaris nearly orthogonal to the capillary plane. This causes blood entering from the arteriole to rapidly dissipate centrifugally towards collecting venules. This dissipation is observed in computational modeling [37, 38] and empirical observation in non-human primates [42, 43], which together suggest that the choriocapillaris exhibits substantial gradients in hemodynamic parameters as a function of distance to arteriolar and venule junctions. Although pending further validation, these findings suggest that shear stresses also vary significantly within the choriocapillaris, based on the relative distance from a feeding arteriole and draining venule. Similar to arterial branch points and curvatures where atherosclerotic plaques are focally deposited, the hemodynamic environment of the choriocapillaris is extremely anatomically heterogeneous. We hypothesize that this heterogeneity renders the choroid, and adjacent RPE and retina that rely upon it, protected from or prone to AMD pathogenesis.

Figure 3. Hemodynamics of the Choriocapillaris.

In this diagram, light is depicted as entering the eye upwards. Blood enters the choriocapillaris through terminal arterioles that form junctions orthogonal to the capillary plane, and exit through venules in much the same manner. As blood exits the arteriole, it spreads centrifugally. This unique anatomic scenario causes a substantial drop in hemodynamic parameters from relatively high pressures, shear stresses, and radiative heat capacity at the arterial inlet, to the relatively low pressures, shear stresses and radiative capacity at the venous exist sites. These parameters therefore are predicted to vary substantially over a relatively short distance. Residence time follows an inverse pattern. The local variability in hemodynamics is reminiscent of arterial branches, where local hemodynamics exhibit strong anatomic dependence.

How Choroidal Hemodynamics Affect Function

Given the apparent heterogeneity of blood velocity in the choriocapillaris measured empirically and as predicted by computational modeling, it is important to understand the influence of local blood flow on choroidal endothelial cell (EC) function. For example, one would expect transport of oxygen, glucose and macromolecules from the choroid to the RPE layer to depend greatly on local blood flow. Computational modeling of choroidal geometry also predicts a strong interrelationship between anatomical parameters and blood residence time [37], a measure of the time which a substance remains within a unit system and a critical parameter in heat dissipation [44]. Another parameter affected by local blood flow is the magnitude of shear stress imparted on the vascular endothelium. Shear stress is a mechanical drag that is linearly related to both the shear rate (the spatial gradient of fluid velocity orthogonal to the vessel wall) and the viscosity of the fluid. As described below, shear stress dramatically influences EC behavior.

Computational modeling of choriocapillary blood flow predicts that local pressure and velocity are inextricably linked to anatomic architecture of the choroidal functional unit [37, 38]. The anatomy of the choroid is fundamentally transformed in aging and in AMD as described below. Therefore, it is critical to understand how anatomic changes in the choroid, as observed in healthy aging or in diseases such as AMD, affect local hemodynamics. For example, choriocapillaris dropout is exacerbated in AMD [18–21]. One would predict that focal choriocapillaris atrophy would increase local vascular resistance. Holding flow rate constant, this implies that capillary atrophy would cause a local increase in both blood velocity (by conservation of mass) and consequently shear stress in the residual capillaries. This assumption is complicated by the observation that bulk measurements of choroidal blood flow have shown significant reductions in severe AMD relative to healthy control human subjects [13]. Thus, precisely how local hemodynamic parameters are affected by choriocapillaris anatomic changes occurring in AMD is currently unknown. Nevertheless, descriptions of the anatomic changes in the choroid during AMD progression have advanced considerably. As described above, recent studies have reported that in GA, the atrophy of the choriocapillaris extends beyond the margin of RPE loss, suggesting that it precedes RPE atrophy in AMD [18–21]. The implication of these findings is that the extent of choriocapillaris atrophy affects hemodynamic parameters such as shear stress and residence time, and consequently, the pathophysiologic impact of these parameters could contribute to AMD progression in other layers such as the RPE and retina. We hypothesize that initial hemodynamic gradients within the choriocapillaris render areas susceptible to atrophy or proliferation. Once the disease begins, anatomic transformation of the choriocapillaris as a consequence of disease causes further hemodynamic alterations which in turn exacerbate disease progression. This concept is analogous to the notion that local hemodynamic changes due to atherosclerotic plaque development are responsible for distal expansion of the plaque.

Molecular and Cellular Consequences of Choroidal Hemodynamics

Although studies directly assessing the role of mechanical forces on choroidal endothelium are lacking, data generated using other vascular beds, or cultured ECs derived from other vascular beds have provided potentially meaningful insights into the molecular mechanisms implicated in AMD pathogenesis. An important caveat is that vascular beds exhibit substantial regional heterogeneity. For example, transcriptional, proteomic, and functional profiling of primary human choroidal and retinal EC have displayed substantial divergence [45–48]. Consequently, hemodynamic responses of commonly utilized ECs (such as human umbilical vein (HUVEC)) need not be conserved in the unique hemodynamic environment and cellular millieu of the choroid. To our knowledge, no published study exists that directly interrogates the role of hemodynamics on choroidal EC biology. Nonetheless, several responses of the EC shear response are relevant for AMD pathogenesis. Broadly, the endothelial shear stress response can be divided into three groups, based on properties of shear stress stimulation: i) early responses to flow initiation, ii) responses to prolonged high-magnitude, unidirectional shear stress, often referred to as atheroprotective shear stress, (with signals imparted to the endothelium lining the large ‘straight portions’ of arteries resistant to atherosclerosis) and iii) responses to prolonged low-magnitude, oscillatory shear stress, or, atheroprone shear stress, (with signals imparted to the endothelium lining ‘branched or curved portions’ of arteries susceptible to atherosclerosis). Another important aspect to the theory is that choroidal anatomy is fundamentally different in the macula compared to the periphery. Whereas the junction of arterioles and venules are almost orthogonal in the macula, venules and arterioles intersect with choriocapillaris at more gradual angles in the periphery. This suggests that the hemodynamic environment of the periphery is more likely to be uniform compared to the highly heterogeneous macula, as observed in non-human primates [42]. Consequently, we predict that the extreme heterogeneity of shear stress in the macula is more uniform in the periphery. It follows that peripheral areas may be less ‘primed’ by lack of exposure to ‘atheroprone’ shear stress, and thus, less susceptible to disease progression.

VEGF Signaling and Angiogenesis

Anti-VEGF therapies have revolutionized the management of neovascular AMD (Box 2). VEGFA is also a trophic factor for both the healthy choroid and the neuroretina, and prolonged use of these therapies can eventually induce macular atrophy. VEGF signaling is an indispensable component of endothelial shear stress responses. VEGFR2 and VEGFR3 are components of a mechanosensory complex which mediates a number of shear stress responses [49, 50]. Acute application of arterial levels of shear stress (12 dyne/cm²) in vitro using a cone and plate viscometer induces rapid VEGFR2 phosphorylation in primary bovine ECs, similar to stimulation with VEGFA [51]. Whether VEGFR2 activation by shear stress depends on ligand binding is controversial. For instance, pretreatment of primary bovine EC with a VEGFA neutralizing antibody does not inhibit flow-induced VEGR2 phosphorylation [51]. Conversely, other studies argue that VEGR2 activation by flow relies instead on a pool of VEGFA normally inaccessible, either bound within the extracellular matrix, or released either by flow stimulation by matrix metalloproteinases [52], or between EC in a juxtacrine manner [53]. In either case, shear stress-dependent VEGF signaling may be recalcitrant to anti-VEGFA therapies utilized clinically. Prolonged application of atheroprotective shear stress increases VEGFA protein expression in HUVEC and primary bovine distal pulmonary microvascular ECs [53–55], enhances EC survival under serum starvation challenge in vitro [53]. Atheroprotective shear stress also significantly blunts primary bovine EC proliferation as measured by cell cycle analysis [56, 57], suggesting that EC residing in the area exposed to this shear stress regime exhibit robust autocrine-juxtacrine VEGFA-dependent cytoprotection against starvation conditions, and proliferative quiescence. By comparison, areas of the vasculature exposed to atheroprone shear stresses that do not confer proliferative quiescence may be particularly susceptible to mitogenic signaling in EC, a process that contributes to angiogenic expansion of the choroidal vascular network [58].

In addition to the effects on resident ECs, circulating angiogenic cells (CACs), formerly referred to as endothelial progenitor cells (EPC), have been thought to contribute to the development and elaboration of the neovascular lesion in wet AMD by providing paracrine support of the aberrant angiogenic resident endothelium. In human CNV tissues and in laser injury-induced CNV in mice, CAC-derived cells account for a significant fraction of cellular content of the neovascular membrane [59–61]. Thus, inhibiting CAC trafficking is a proposed therapeutic strategy [60, 61]. Because, atheroprone shear stress is an important stimulus for CAC trafficking into the vascular endothelium [62], reduced hemodynamic forces could potentially promote CAC recruitment. Shear stress promotes CAC differentiation [63, 64], although it is not known whether CAC subtypes implicated in neovascular AMD are sensitive to shear stress. Collectively, we propose that atheroprone shear stress in the choriocapillaris is a critical determinant of angiogenesis through the activation and recruitment of multiple cell types that collaborate to promote aberrant neovascularization.

Immune Signaling, Cell Trafficking and Inflammation

Shear stress strongly influences diverse endothelial inflammatory responses, including those associated with AMD. In vitro application of atheroprotective shear stress induces expression of the transcription factors Kruppel like factor 2 (KLF2) and nuclear factor, erythroid 2 like 2 (NFE2L2) whose transcriptional targets suppress NF-κB activation in a variety of cell culture model systems [65]. Further, expression of these transcription factors localizes to regions of the arterial tree exposed to atheroprotective shear stress in vivo, in mouse and human studies. Conversely, atheroprone shear stress in vitro in animal models, and in vivo in primary human aortic ECs, induces a variety of signals including induction of inflammatory transcription factor NF-κB [66], and the danger sensor Receptor of Advanced Glycation Endproducts (RAGE) [67]. These signaling pathways, among others, sensitize ECs exposed to atheroprone shear stress to endothelial barrier dysfunction and inflammation.

Atheroprone shear stress in HUVEC also induces NLRP3 inflammasome activation [68], an inflammatory pathway implicated in AMD pathogenesis in RPE cell cultures and rodent models, as well as analyses of postmortem human eyes [69–73]. Of note, NLRP3 inflammasome activation in the choroid has been detected in human AMD donor eyes [72].

Human eyes with intermediate and advanced AMD also exhibit robust increases in macrophage content [74, 75], and macrophage depletion significantly reduces CNV in a mouse model of laser injury [76]. Atheroprone shear stress also increases endothelial-monocyte adhesion to HUVEC [77] and mouse aortic EC [78]. Furtermore, shear stress is a strong predictor of macrophage extravasation in the aorta of atherosclerosis prone mice [79], suggesting the hemodynamic environment is an important regulator of immune cell trafficking.

The most thoroughly studied immune pathway in AMD is the complement cascade, first observed in blood vessels of surgically excised neovascular AMD membranes [80] and subsequently popularized by genome-wide association studies, which report that polymorphisms in several complement related genes confer statistical risk for AMD development (reviewed in [81]). Activation of the complement membrane attack complex (MAC) is strongest in the choriocapillaris of AMD donor eyes relative to age-matched healthy controls [18]. A cell culture study of a rhesus chorioretinal EC line suggests that MAC deposition may contribute to AMD pathogenesis via induction of characteristic phenotypes of choroidal endothelium in AMD, namely, a concomitant atrophy and proliferation [82]. CD59 is an important negative regulator of complement activation, which in mice protects the RPE/choroid from inflammatory damage [83] and may represent a putative therapeutic target for AMD [84, 85]. In vitro atheroprotective, (but not atheroprone) shear stress induces CD59 expression in HUVEC, and CD59 is upregulated in atheroprotected areas of mouse aorta [86]. Collectively, these findings indicate that areas of the choriocapillaris exposed to atheroprone shear stress may be more susceptible to inflammatory activation. However, further experimental validation will be required to fully establish these associations.

Vascular Barrier Function

Fluid leakage into the retina from aberrant neovessels is a clinically relevant outcome of neovascular AMD [87]. Barrier function within the healthy choriocapillaris is unique. The apical surface of the choriocapillaris is fenestrated, allowing the free transport of large macromolecules through passive diffusion. Little is known about the effects of shear stress on EC fenestrations. However, whether neovascular lesions are indolent or active is thought to depend on stability of EC-cell-cell junctions. Shear stress causes dynamic remodeling of the intercellular junctions, including reorganization of the adherens junction [88–90], and eventually the formation of stable tight barriers to facilitate transport [91, 92]. Prolonged exposure to atheroprone shear stress induces activation of p21-activated kinase which controls endothelial cell-cell junction integrity [93, 94], leading to impaired trans-endothelial resistance [94]. These findings imply that atheroprone shear stress in the developing neovascular lesion may contribute to impaired barrier function.

Overall, a better understanding of the hemodynamic environment within the degenerating choroid and the neovascular lesion, and its influence on choroidal EC behavior could shed light on endothelial transport function in AMD.

Heat Dissipation

One role of the choroid is to act as a cooling radiator to remove excess heat which accumulates due to absorption of focused light. Reduced choroidal perfusion decreases its heat radiating capacity in an experimental model of intraocular pressure induced choroidal flow impairment [95] and in a computational model [96]. Moreover, production of VEGFA by the RPE is temperature dependent [97–99], suggesting that an indirect effect of reduced choroidal perfusion in AMD contributes to heat-induced VEGFA secretion. Another potential deleterious consequence of decreased heat dissipation by the choroid is induction of Alu RNAs, endogenous activators of the NLPR3 inflammasome which have been implicated in GA in human AMD specimens, mouse models and in primary cell cultures [73, 100, 101]. Alu RNAs and analogous rodent transcripts are upregulated by heat stress in HEK293, HeLa, K562, and 3T3 cells [102, 103]. Conversely, in a mouse laser injury model of CNV, heat preconditioning promotes anti-angiogenic matrix deposition in RPE cells, thus reducing neovascularization [104]. Heat pre-treatment of RPE cells confers protection against oxidative damage, blunts oxidative stress-induced VEGFA secretion [105], and enhances heat shock protein 70 expression [106]. Therefore at the onset of AMD pathology, the loss of the choroid’s capacity to dissipate retinal heat could modulate the immunologic activation and angiogenic capacity as the disease progresses.

Concluding Remarks - A revised Model of Choroidal Hemodynamic Dysfunction in AMD

We propose two major revisions to Friedman’s hemodynamic model (Figure 4, Key Figure). The first is that choroidal hemodynamics contribute to AMD pathogenesis in a substantially more localized manner than has been appreciated. Given emerging empirical and computational evidence of significant heterogeneity of hemodynamics in the choriocapillaris, we hypothesize that hemodynamic parameters within individual capillaries are pivotal factors which promote or prevent AMD pathogenesis. For example, based on observations from other vascular beds, we envisage that local reductions in hemodynamic shear stress promotes (or licenses) choroidal proliferation. By comparison, local elevations in shear stress are predicted to prevent choroidal EC remodeling, immune cell trafficking and transcellular transport. Our theory suggests that local hemodynamics, which likely vary on the scale of a single capillary, determine whether the choroidal endothelium undergoes apoptosis or proliferation under conditions of environmental, genetic or systemic stress. In this way, the present theory may combine with other proposed disease mechanisms by providing an explanation for the focal manifestation of systemic AMD risk contributors such as genetics, smoking, and comorbid cardiovascular risk factors. The second major revision we propose is that the consequences of altered choroidal perfusion extend beyond hypoxia and passive transport of waste from the RPE. Instead, we posit that hemodynamic parameters themselves (e.g. endothelial shear stress, residence time, heat dissipation) comprise fundamental micro-environmental cues that transform the molecular trajectory of the choroid, and indirectly, the RPE and outer retina. Clinicians have long recognized the value of quantifying choroidal blood flow in the diagnosis and prognosis of this disease (Box 3). New technologies and detailed mechanistic studies have provided a more resolved picture of the choroidal landscape of AMD pathogenesis, although many questions and further validation of the proposed model still remain (see Oustanding Questions). Now is the time to bring the power of molecular biology to confirm or refute decades of old hypotheses of choroidal hemodynamics as an underlying cause of AMD.

Key Figure, Figure 4. A Revised Hemodynamic Model of AMD.

The model posits that the normal, healthy choriocapillaris consists of resistant (red) and susceptible (brown) areas, based on the local hemodynamic environment. Relevant factors in addition to those mentioned by Friedman include distance from the terminal arteriole, density of arterioles and venules, as well as choriocapillaris thickness and density. With aging, susceptible areas may eventually adopt an atrophic phenotype (beige), which alters local hemodynamics. By virtue of decreased CC density, atrophied areas lose heat-radiating capacity, thereby increasing inflammatory mediators such as complement and Alu RNA. In areas of the CC, atrophy induces elevated shear stress in adjacent capillaries, rendering the choroid ill-suited to remove retinal waste and incapable of vascular repair. This ultimately leads to GA by choroidal RPE and retinal atrophy. Other areas of the CC experience reduced shear stress, thereby promoting an ‘activated’ endothelial phenotype. The angiogenic and inflammatory influences of low shear stress and hyperthermia of the RPE conspire to produce a choroidal neovascular membrane.

Box 3. Clinician’s Corner.

• The development of age-related macular degeneration (AMD) is strongly associated with hemodynamic abnormalities in the choroidal vasculature. However, measurements of the hemodynamic changes preceding AMD are poorly resolved with respect to the discrete appearance of disease manifestations (e.g. drusen, choriocapillaris dropout).

• Hemodynamic forces such as endothelial wall shear stress are important pathologic parameters predicting the development of vascular diseases such as atherosclerosis as well as aneurysms. Shear stress is an important mediator of endothelial cell health.

• A newer model of AMD pathogenesis emerges, positing that local hemodynamic aberrations within the choriocapillaris could be causative factors contributing to the pathogenesis of AMD.

• In the future, delineating the relationship between choriocapillaris hemodynamics might provide new tools to predict areas that may be susceptible to AMD development, as well as new insights into altered pathways stemming from aberrant hemodynamics and leading to putative molecular targets to treat this disease.

Trends.

• The choroid is an anatomically unique vascular bed that provides essential trophic support for the retina.

• Alterations in choroidal anatomy and hemodynamics are strongly associated with development of age-related macular degeneration (AMD), the leading cause of blindness in the industrialized world.

• Rather than passive conduits to flow, the vascular endothelium is highly active, and strongly influenced by local blood flow.

• A variety of AMD-associated signaling pathways are hemodynamically regulated in the vascular endothelium.

• A revised model of choroidal hemodynamics as an underlying cause of AMD pathogenesis is emerging, wherein local hemodynamic parameters influence choroidal and retinal pigmented epithelial cell health.

OUTSTANDING QUESTIONS BOX.

• How does the choriocapillaris endothelium shear stress response (SSR) affect its surrounding microenvironment such as Bruch’s membrane and the RPE layer?

  • Studies of endothelial cells (EC) in other vascular beds provide insight into the SSR conserved between diverse tissue areas. How the choroidal EC SSR influences surrounding cell types such as the RPE and resident non-endothelial choroidal cells, and structures such as Bruch’s membrane is unknown.

• How does the RPE and surrounding choroidal microenvironment modulate the choroidal EC SSR?

  • How the choroidal EC SSR is modulated by unique cellular and molecular constituents comprising the choroid and RPE is unknown.

• In what ways does anatomic transformation of the choroid influence local hemodynamics forces and parameters?

  • Simplified anatomic models of healthy choriocapillaris demonstrate substantial regional heterogeneity of hemodynamic parameters. Understanding how these parameters are affected by choroidal neovascularization and atrophy is critical in determining how the hemodynamic response may aggravate or mitigate disease progression.

• Can choriocapillaris hemodynamics be utilized to predict localized AMD risk?

  • The focal development of atherosclerosis is predictable if given a thorough hemodynamic description. Can ever-improving ocular imaging modalities and computational modeling be utilized to provide a spatially resolved risk profile for the development of AMD pathologies?

• Can correcting choroidal hemodynamics represent a therapeutic strategy for AMD?

  • Given the association of hemodynamic abnormalities with AMD pathogenesis, are there surgical or pharmacologic strategies that can be deployed to prevent, delay or reverse the progression of AMD? Alternatively, can the deleterious effects of the abnormal hemodynamic environment be circumvented by molecular targeting of the choroidal endothelium to treat AMD?

Glossary of terms

Macula

specialized area of the retina responsible for color vision and fine visual acuity. The vascular supply and retinal cells are unique in this area, reflective of its extreme metabolic demand.

Choroid

One of two blood supplies nourishing the retina. Whereas the retinal vasculature supplies the inner retinal layers, the choroid provides the majority of nutrients required by the outer retina (photoreceptors), and RPE. The choroid is divided into three layers: Outermost, includes Haller’s layer of large arterioles and venules; medial or Sattler’s layer; and innermost, the choriocapillaris.

Sattler’s layer

layer of the choroid which is comprised of medium-sized arterioles and venules. These vessels supply blood to the choriocapillaris.

Choriocapillaris

innermost layer of the choroid, comprised of a thin single layer of fenestrated capillaries. The capillary bed is dense and highly interconnected (anastomosed).

Bruch’s membrane

thick acellular membrane lying between the choriocapillaris and the RPE. Ruptures in Bruch’s membrane are a critical step in choroidal neovascularization.

Drusen

Extracellular deposits mostly accumulating within (or adjacent to) Bruch’s membrane. Subretinal drusenoid deposits are a subtype forming between the RPE and photoreceptor layers.; Drusen are the first clinical sign of AMD, and their biogenesis and reabsorption are thought to be both a sign and symptom of age-related pathology.

Geographic atrophy (GA)

The advanced and irreversible binding form of ‘dry’ AMD demarked by confluent areas of macular RPE atrophy.

Choroidal neovascularization (CNV)

pathological growth of aberrant blood vessels from the choroid, across Bruch’s membrane and into the sub-RPE or subretinal space. Leakage of CNV vessels is the most common cause of vision loss in neovascular AMD, and occurs in a number of visual diseases such as myopia.

Neovascular age-related macular degeneration

the most prevalent advanced form of AMD characterized by leakage blood constituents from CNV lesions (or occasionally aberrant growth of retinal vessels). Neovascular AMD is managed by anti-VEGF therapies, but significant healthcare challenges remain.

Stargardt disease

inherited juvenile macular degeneration, whose clinical manifestation bear similarities to atrophic AMD. Most commonly, Stargardt disease is caused by a mutation in the ABCA4 gene, which causes degeneration of the RPE.

Optical coherence tomography angiography (OCTA)

non-invasive imaging technique to visualize the retinal and choroidal circulation. Recent advances in OCTA have made observations of choroidal vascular changes in AMD progression easier, more quantitative, and better resolved than previously possible.

Exudate

Fluid leakage from aberrant neovessels disrupting the retinal architecture. Exudate is strongly and negatively correlated with visual acuity.

Hemodynamics

fluid mechanics of blood flow. Parameters include pressure, velocity, flow rate, viscosity, shear rate and shear stress. A hemodynamic description of blood flow also provides information about how the fluid interacts with the tissue environment, such as in terms of nutrient, waste and heat transport.

Shear stress

frictional drag that a moving fluid exerts on a solid interface. In the context of the circulatory system, as blood flows across a vessel wall it exerts shear stress on the endothelium. Endothelial cells are sensitive to their shear stress environment.

Atheroprotective and atheroprone shear stress

Types of shear stress patterns observed in regions resistant to or predisposed to atherosclerosis development respectively. Atheroprotective shear stress is high magnitude, unidirectional, and pulsatile. Atheroprone flow is low magnitude and changes directions.

Vascular endothelial growth factor A (VEGFA)

promotes the survival and development of new blood vessels. Extracellular VEGFA also reduces cell-cell integrity which increases the exudation of choroidal neovessels.

Anti-VEGF therapies

compounds that prevent extracellular VEGFA from signaling. Three such therapies are approved for the treatment of neovascular AMD.

NLRP3 inflammasome

innate immune signaling complex whose activation is implicated in the pathogenesis of AMD. It is activated by a variety of AMD-associated stressors including Alu RNA, amyloid β, and cigarette smoke. NLRP3 inflammasome activation causes secretion of inflammatory cytokines IL-1β and IL-18, both thought to be involved in AMD pathogenesis.

Retinal pigmented epithelium (RPE)

single cell layer situated between the photoreceptors and Bruch’s membrane. It provides trophic support for the choroid and photoreceptor layer. Degeneration of the RPE layer is a hallmark of GA.

Photoreceptors

Specialized neurons that convert light into electrochemical signals necessary for vision. The two major types are rods, sensitive to low levels of light, and cones, capable of distinguishing wavelengths (colors) of light.

Outer segments

Processes of the photoreceptors in which light is absorbed. Constituents of outer segments are periodically shed and recycled by the RPE to support continuous rejuvenation of the phototransduction machinery.

Sclera

white fibrous, protective coating of the eye which is situated posterior to the choroid. Scleral rigidity is implicated in AMD pathogenesis.

Hypoxic response

coordinated cellular response to conditions of poor or no oxygenation. One such adaptive response is the development of new blood vessels to restore transport of oxygen and metabolic nutrients.

Watershed zones

Anatomic areas whose vascular supply is at the extreme distal end of two or more feeding arteries. These areas are thought to be susceptible to hypoperperfusion; tissue metabolism requires full perfusion of both feeding arteries as the vascular networks that branch from the arteries are non-redundant.

Posterior pole

posterior hemisphere of the eye surrounding the macula, encircled by the periphery. The choroidal vasculature in the posterior pole is distinct from both the macula and the periphery by presenting a greater density of feeding arterioles and draining venules.

Neuroretina

refers to the neuronal layers that comprise the retina, excluding the RPE and blood vessels.

Membrane attack complex (MAC)

multiprotein complex that is a terminal effector of the complement cascade activation. Deposition of MAC on cells surfaces causes cell lysis, and in submaximal conditions (sublytic) conditions, induces cell inflammation and dysfunction. MAC deposition in choroidal endothelial cells is one proposed mechanism of choroidal atrophy in AMD.

Adherens junction

type of cell-cell junction thought to promote monolayer maturation and integrity, as well as serve as a signaling hub for the endothelial shear stress response.

Alu RNAs

Long non-coding RNAs derived from Alu repetitive genetic elements, ubiquitous in the human genome. Accumulation of Alu RNA is implicated in inflammatory-mediated cell death in AMD.