Current Concepts in the Pathogenesis of Age-Related Macular Degeneration

Objective  To review and synthesize information concerning the pathogenesis ofage-related macular degeneration (AMD).

Methods  Review of the English-language literature.

Results  Five concepts relevant to the cell biology of AMD are as follows: (1)AMD involves aging changes plus additional pathological changes (ie, AMD isnot just an aging change); (2) in aging and AMD, oxidative stress causes retinalpigment epithelial (RPE) and, possibly, choriocapillaris injury; (3) in AMD(and perhaps in aging), RPE and, possibly, choriocapillaris injury resultsin a chronic inflammatory response within the Bruch membrane and the choroid;(4) in AMD, RPE and, possibly, choriocapillaris injury and inflammation leadto formation of an abnormal extracellular matrix (ECM), which causes altereddiffusion of nutrients to the retina and RPE, possibly precipitating furtherRPE and retinal damage; and (5) the abnormal ECM results in altered RPE-choriocapillarisbehavior leading ultimately to atrophy of the retina, RPE, and choriocapillarisand/or choroidal new vessel growth. In this sequence of events, both the environmentand multiple genes can alter a patient's susceptibility to AMD. Implicit inthis characterization of AMD pathogenesis is the concept that there is linearprogression from one stage of the disease to the next. This assumption maybe incorrect, and different biochemical pathways leading to geographic atrophyand/or choroidal new vessels may operate simultaneously.

Conclusions  Better knowledge of AMD cell biology will lead to better treatmentsfor AMD at all stages of the disease. Many unanswered questions regardingAMD pathogenesis remain. Multiple animal models and in vitro models of specificaspects of AMD are needed to make rapid progress in developing effective therapiesfor different stages of the disease.

Age-related macular degeneration (AMD) is the leading cause of blindnessand visual disability in patients 60 years or older in the Western hemisphere.¹ The clinical presentation of AMD includes drusen,hyperplasia of the retinal pigment epithelium (RPE), geographic atrophy, andchoroidal new vessels (CNVs).² Only approximately10% to 15% of patients with AMD have severe central vision loss. AtrophicAMD, characterized by outer retinal and RPE atrophy and subjacent choriocapillarisdegeneration, accounts for approximately 25% of cases with severe centralvision loss.¹ Exudative AMD is characterizedby CNV growth under the RPE and retina, with subsequent hemorrhage, exudativeretinal detachment, disciform scarring, and retinal atrophy. Serous or hemorrhagicpigment epithelial detachment also occurs. Exudative AMD accounts for approximately75% of cases with severe central vision loss.¹ Mostpatients with subfoveal choroidal neovascularization develop profound centralvision loss regardless of whether the CNV has classic or occult morphologicfeatures on angiography.^3,4 Rarely,patients have peripheral and central vision loss due to extensive subretinaland vitreous hemorrhage.

The prevalence of early AMD (ie, the presence of soft indistinct orreticular drusen or drusen with RPE degeneration or hyperpigmentation) is18% in the population aged 65 to 74 years and 30% in the population olderthan 74 years.⁵ Findings from 2 population-basedstudies^6,7 indicate that the prevalenceof geographic atrophy is 3.5% in persons older than 75 years, or approximatelyhalf the prevalence of CNVs. Since the population older than 65 years is thefastest growing segment of our society, the burden of disease will increaseduring the 21st century.⁸ Considering the highsocial and financial cost of this problem, the need for new therapies to preventand treat exudative and atrophic maculopathy is pressing. Many different strategiesare being pursued, ranging from antiangiogenic therapy to transplantationsurgery. The purpose of this review is to summarize recently developed experimentaland clinical biological data relevant to the pathogenesis of AMD.

Aging is associated with biological changes in the eye. These featuresof aging are present in AMD eyes and may contribute to the pathogenesis ofAMD, but they do not lead inevitably to AMD. Thus, it is important to recognizeaging changes in the RPE–Bruch membrane–choriocapillaris complexthat occur in aged eyes without AMD.

In general, aging is associated with cumulative oxidative injury.⁹ For example, postmitotic cells such as RPE cells accumulatemitochondrial DNA deletions and rearrangements with aging.¹⁰ Verzar¹¹ recognized that aging is associated with extracellularmatrix (ECM) alterations. These may include abnormalities of ECM biosynthesis;postsynthetic modifications of ECM, including degradation; altered interactionamong ECM components; and changes in cell-ECM adhesion.¹² Agedhuman fibroblasts, for example, seem to produce structurally and functionallyabnormal fibronectin that exhibits reduced binding to native types I and IIcollagen.¹³ Changes in the extracellular environmentcan induce changes in the cell phenotype.¹² Manyof these changes may be under genetic control. Fibroblasts from patients withWerner syndrome, for example, exhibit each of these abnormalities.^14-16 (Werner syndromeis a condition associated with premature aging that results from a loss offunction mutations in the WRN gene [which encodesa DNA helicase], which leads to rapid telomere shortening.¹⁷)Epigenetic reactions involved in aging include the Maillard reaction, uncontrolledproteolytic degradation, and free radical release. The Maillard reaction isthe reaction of free reducing sugars or reactive aldehydes with free aminogroups to form Schiff bases and, after Amadori rearrangements, polycyclicadvanced glycation end products.¹⁸ Advancedglycation end products induce cell injury (directly or through cell surfacereceptors) and can induce dysregulation of tissue remodeling with enhanceddeposition of ECM. Each of these features of aging is relevant when consideringthe aging of the retina–RPE–Bruch membrane–choriocapillariscomplex and the pathogenesis of AMD (Figure1).

Lipofuscin comprises a group of autofluorescent lipid-protein aggregatespresent in nonneuronal and neuronal tissues. As is the case for many postmitoticcells, lipofuscin accumulates in RPE cells during life. In one study,¹⁹ lipofuscin occupied 1% of the RPE cytoplasmic volumeduring the first decade of life and 19% of the cytoplasmic volume by age 80years. Reduction in functional cytoplasmic volume might compromise RPE function,for example, phagocytosis,²⁰ which can leadto photoreceptor death. In the RPE, the major source of lipofuscin is theundegradable products of photoreceptor outer segment metabolism.²¹ Intralysosomaliron–catalyzed reactions generate lipofuscin. By producing reactiveoxygen species, lipofuscin may induce oxidative damage in the RPE and surroundingtissues and may inhibit RPE lysosomal enzyme activity (see the bulleted list).Okubo and coworkers²² found a linear relationshipbetween RPE autofluorescence and Bruch membrane thickness, which indicatesthat aging changes in the RPE and Bruch membrane may be related.

Bruch membrane thickness seems to increase linearly with aging fromapproximately 2 µm at birth to approximately 4 to 6 µm in thetenth decade of life.²³ Bruch membrane thickeningcan arise from increased production and decreased degradation of extracellularmaterial. As noted in the "Bruch Membrane Composition and Permeability" subsection,changes in thickness are associated with changes in protein composition, proteincross-linking, increased glycosaminoglycan size, and increased lipid content.Age-related thickening of the Bruch membrane is not confined to the innercollagenous layer. For example, native Bruch membrane collagen content increasesin the outer collagenous layer during the teens. By age 40 years, wide-spacedcollagen also accumulates in this layer.²⁴ Periodicacid–Schiff–positive material that resembles the contents of RPEphagosomes accumulates in the inner collagenous layer and, later, in the elasticlayer.²⁵ Thus, during aging, dysfunctionalRPE cells might produce abnormal quantities of ECM material, including cellfragments, collagen, and other basement membrane components.^26,27

Impaired ability to degrade the ECM might contribute to age-relatedBruch membrane thickening. Matrix metalloproteinases (MMPs), for example,are zinc-dependent enzymes that catabolize ECM proteins, including collagenand elastin. Tissue inhibitors of metalloproteinases (TIMPs) regulate theactivity of MMPs. Retinal pigment epithelium cells produce MMPs and TIMP-3.^28-30 Inactive forms ofMMP-2 and MMP-9 increase in the Bruch membrane with aging, particularly inthe submacular Bruch membrane.³¹ Abnormalitiesin metalloproteinase activity can result in changes in Bruch membrane thickness.Mutations of TIMP-3, for example, can cause decreased TIMP-3 turnover andresult in Sorsby fundus dystrophy, a condition characterized by the accumulationof abnormal extracellular material between the RPE and the inner collagenouslayer of the Bruch membrane.^32,33 Bruchmembrane TIMP-3 content increases with age.³⁴ SinceTIMP-3 gene expression in the macular area does not increase substantiallywith aging,³⁵ there may be altered TIMP-3 turnoverwith sustained MMP inhibition during aging.

Age-related declines in choriocapillaris density and lumen diameter(see the "Choroidal Blood Flow" subsection) might also decrease clearanceof debris from the Bruch membrane, which would contribute to thickening withage. In Sorsby fundus dystrophy, symptoms and retinal function improve withhigh-dose vitamin A therapy, suggesting that impaired diffusion across theBruch membrane can be a consequence of Bruch membrane thickening.³⁶ (The material that accumulates in the Bruch membranein this condition resembles the abnormal extracellular debris deposited inthe Bruch membrane in AMD.)

Various collagens, glycosaminoglycans, laminin, and fibronectin arenormal constituents of the Bruch membrane (Table 1). With aging, the Bruch membrane thickens, periodic acid–Schiffstaining increases, and Bruch membrane type I collagen increases.^26,35,39-42 Membranousdebris, filamentous material, and coated vesicles accumulate primarily inthe inner collagenous layer by early adulthood and continue to do so throughoutadult life.³⁹ With aging, collagen cross-linkingseems to increase in the Bruch membrane, and there is a significant increasein the amount of noncollagen protein in the submacular Bruch membrane butnot in the periphery, which might mean that protein-containing debris is trappedin the Bruch membrane during aging.⁴³ By latemiddle age, lipid deposition in the Bruch membrane is apparent.³⁹ Basallaminar deposit, which comprises mostly wide-spaced collagen⁴⁰ andother materials, including laminin, membrane-bound vesicles, and fibronectin,is present in the seventh decade of life during normal aging.^44-46 Pauleikhoffand coworkers⁴⁷ reported an age-related declinein the presence of laminin, fibronectin, and type IV collagen in the RPE basementmembrane (especially over drusen). Basal linear deposit, consisting primarilyof granular and vesicular material with foci of wide-spaced collagen, appearsin older persons and is more specific for AMD.^44,48 Duringaging, Bruch membrane glycosaminoglycans increase in size, and heparan sulfatecontent increases.³⁸ Advanced glycation endproducts accumulate in the Bruch membrane during aging.⁴⁹ Advancedglycation end products have been shown to promote trapping of macromolecules,^50,51 and they might alter cellular traffickingthrough the Bruch membrane, particularly if the cells express receptors foradvanced glycation end products. Molecules present in the Bruch membrane imparta negative electrostatic charge at physiologic pH.^38,52 Age-relatedchanges in glycosaminoglycans might alter this charge and, as a result, thepermeability properties of the Bruch membrane.³⁸

Thus, the molecular composition of the Bruch membrane and the tightjunctions between RPE cells affect the movement of molecules between the choriocapillarisand the subretinal space. Most evidence indicates that the hydraulic conductivityof the Bruch membrane decreases exponentially with age in healthy individuals.^41,53 At any given age, the submacularBruch membrane is affected to a greater degree than the peripheral Bruch membrane.Starita and coworkers⁴² used excimer laserablation of different layers of the Bruch membrane to demonstrate that mostof the resistance to water flow lies in the inner collagenous layer of theBruch membrane. These investigators suggested that a high-resistance barrierdevelops in the Bruch membrane in older eyes, probably due to lipid and vesicular-granulardebris entrapment in the Bruch membrane.

Most of the Bruch membrane hydraulic conductivity decrease occurs byage 40 years. Marshall and coworkers³⁹ notedthe discrepancy between the early rapid decline in conductivity and the relativelyslower rate of increase in Bruch membrane thickness. Age-related changes inBruch membrane biochemical composition probably underlie the discrepancy.Specifically, there is increased lipidization, protein cross-linking, andprotein deposition in the Bruch membrane with aging. Lipid accumulation inthe Bruch membrane begins to increase substantially after age 40 years.^54,55 The rate of lipid accumulation underthe macula may be higher than under the peripheral retina, perhaps due tothe greater density of photoreceptors in the macula and a greater susceptibilityof outer segment lipids to peroxidation in the posterior pole. Spaide andcoworkers⁵⁶ found that the amount of peroxidizedlipids in the Bruch membrane increased exponentially with age. The lipidsseemed to be derived from long-chain polyunsaturated fatty acids normallyfound in the outer segments, for example, docosahexanoic acid and linoleicacid, providing support for the notion that at least some of the lipid inthe Bruch membrane is of cellular origin rather than derived from the blood.

Bruch membrane morphometry indicates that the elastin layer has thehighest porosity and that the inner collagenous layer has the lowest porosity.³⁹ The elastin layer seems to become increasingly porouswith age.⁵⁷ This layer might normally constitutea barrier to vessel growth between the choroid and the sub-RPE space, andthis age-related change might have a permissive effect on CNV growth. Marshalland coworkers³⁹ proposed that from the lateteens to the late thirties, membranous debris, vesicles, and collagen accumulationcause a reduction in effective pore size in the inner collagenous layer. Fromthe forties to the sixties, this process continues, and, abetted by substantiallipid deposition, there is an accelerated decline in hydraulic conductivity.At older ages, the deposition of basal laminar and linear deposit furtherreduces functional pore size. In older persons, diffusion of small and largemolecules across the Bruch membrane is impaired.^58-60 Changesin protein cross-linking, noncollagenous protein deposition, and age-relatedlipid accumulation in the Bruch membrane may be the underlying cause.⁶¹ To the degree that hydrodynamic forces alter moleculartransport across the Bruch membrane, it seems possible that hypertension wouldexacerbate age-related trans–Bruch membrane transport problems.

Changes in choroidal blood flow in aging and AMD have been reviewedby Lutty and coworkers.⁶² Ramrattan and coworkers²³ showed that there is a progressive decrease in thethickness of the choroid from 200 µm at birth to 80 µm by age90 years. The choriocapillaris density and lumen diameter decrease, and thewidth of the intercapillary pillars increases with age.^23,63 Inview of these histologic changes, it is not surprising that subfoveolar choroidalblood flow decreases with age.⁶⁴ Indocyaninegreen choriocapillaris filling, for example, is delayed in persons older than50 years, and areas of hypofluorescence are present in the macula of patientswith AMD. Laser Doppler flowmetry of the submacular choriocapillaris demonstratesdecreased choroidal blood flow and volume in individuals older than 46 years,with further reduction in patients with AMD. Guymer and coworkers⁶⁵ pointed out that if choriocapillary endothelial cellprocesses, which are present in the Bruch membrane, play a role in clearingdebris from the Bruch membrane, then an age-related loss of choriocapillariescould play a causal role in Bruch membrane thickening during aging. Alternatively,as the RPE produces substances that help maintain normal choriocapillary densityand anatomy,⁶⁶ Bruch membrane thickening mightcause age-related choriocapillary changes by impairing diffusion of thesesubstances to the choriocapillaris.

Aging is associated with increased oxidative damage.^9,67 Plasmaglutathione levels decrease, and oxidized glutathione levels increase, forexample, with age.⁶⁸ Plasma levels of vitaminC and vitamin E also decrease with age.^69,70 Lipidperoxidation seems to increase with aging.^71,72 Thesusceptibility of RPE cells to oxidative damage increases with aging. Forexample, RPE cell vitamin E levels and catalase activity decrease with aging.^73,74 Macular pigment optical density decreaseswith aging.⁷⁵ Retinal pigment epithelium celllipofuscin content, which enhances susceptibility to oxidative damage, increaseswith aging. In addition, RPE cells that experience phototoxicity exhibit membraneblebbing,⁷⁶ a phenomenon observed in agingand AMD eyes (see the "Evidence of Oxidative Damage in AMD" and "Inflammation"subsections). One study⁷⁷ reported that RPEdensity decreases approximately 0.3% per year throughout life.

Oxidative damage to the RPE is a potential final common pathway forage-related retinal damage that depends on genetic predisposition, cumulativelight damage, free radical injury, and hemodynamic abnormalities (reviewedby Winkler,⁷⁸ Beatty,⁷⁵ andCai⁷⁹ and their colleagues). Production ofreactive oxygen species is stimulated by irradiation, aging, inflammation,increased partial pressure of oxygen, air pollutants, cigarette smoke, andreperfusion injury. Oxygen-derived metabolites cause oxidative damage to cytoplasmicand nuclear elements of cells and cause changes in the ECM. Reactive oxygenspecies react, for example, with nucleic acids, membrane lipids, surface proteins,and integral glycoproteins.

Beatty and coworkers⁷⁵ reviewed the factorspromoting reactive oxygen species formation in the retina and RPE:

• Outer segments are enriched in polyunsaturated fatty acids

• Oxygen tension in the photoreceptor-RPE area is close to thatof arterial blood

• The retina is exposed to high levels of cumulative irradiation

• The retina and RPE contain photosensitizers: rhodopsin, lipofuscin,and cytochrome c oxidase

• The choriocapillaris contains blood-borne photosensitizers

• RPE phagocytosis is an oxidative stress

Briefly, photoreceptor outer segments are enriched in polyunsaturatedfatty acids, which can undergo lipid peroxidation. Lipid peroxidation is greatestin the macula and increases with age.⁸⁰ Invitro evidence^20,81,82 indicatesthat RPE lipofuscin is a photoinducible generator of reactive oxygen speciesthat can compromise lysosomal integrity, induce lipid peroxidation, reducephagocytic capacity, and cause RPE cell death. Lipofuscin granules are continuouslyexposed to visible light and high oxygen tension, which cause reactive oxygenspecies production and possibly further oxidative damage to the RPE cell proteinsand lipid membranes.^78,83 Retinalpigment epithelium lipofuscin is derived in part from vitamin A metabolitesand lipid peroxides.⁸⁴ (Vitamin A is a majorconstituent of photoreceptor outer segments.) The reaction product of ethanolamineand 2 retinaldehyde molecules, N-retinylidene-N-retinylethanolamine (A2-E), is the major photosensitizingchromophore in lipofuscin that causes reactive oxygen species production;A2-E also raises lysosomal pH, thus interfering with lysosomal enzyme activityand reducing lysosomal protein and glycosaminoglycan degradation.^82,85,86 When RPE cells areexposed to light, A2-E conjugated to low-density lipoprotein, which accumulatesin RPE lysosomes, causes loss of lysosomal integrity⁸²;A2-E also inhibits RPE phagolysosomal degradation of photoreceptor phospholipidin vitro.⁸⁷ Retinal pigment epithelium cellswith excessive A2-E exhibit membrane blebbing and extrusion of cytoplasmicmaterial into the Bruch membrane.

One biochemical study⁸⁸ of drusen compositionfound that up to 65% of the proteins identified in drusen are present in drusenderived from AMD as well as healthy age-matched donors. Approximately 33%of the drusen-derived proteins from AMD donors were not observed in healthydonor drusen. These findings may mean that although there is some degree ofcontinuity between aging changes in the Bruch membrane and aging changes associatedwith AMD, there also are distinct differences. For example, in this study,docosahexaenoate lipid–derived oxidative modifications were much morecommon in AMD eyes than in age-matched control eyes. Docosahexanoic acid isa highly unsaturated fatty acid that makes up approximately 50% of rod phospholipids.

One model of aging vs AMD consistent with published clinical, pathological,and experimental observations is shown in Figure 2. Age-related macular degeneration involves aging changesplus additional pathological changes. In AMD and aging, oxidative stress resultsin RPE and, possibly, choriocapillaris injury. In AMD, and possibly in aging,RPE injury elicits an inflammatory response in the Bruch membrane and thechoroid. In AMD eyes, RPE injury and inflammation foster the production ofan abnormal ECM derived largely from the RPE and photoreceptor cells but alsofrom cells in the choroid and from substances in the systemic circulation.The abnormal ECM, in turn, results in altered RPE biologic behavior and maycause further damage to the retina, RPE, and choroid. Oxidative damage tothe choriocapillaris also may contribute to the pathogenesis of AMD. The factorsmediating CNV growth and the development of geographic atrophy involve perturbationof RPE-choriocapillaris homeostasis. Retinal pigment epithelium death, forexample, probably is the cause of choriocapillaris loss in geographic atrophy.^66,89 Evidence for the pathogenic roleof oxidative stress, inflammation, ECM abnormalities, altered RPE biologicbehavior, and genetics is considered in the following subsections.

Clinical Studies of Antioxidants. The Age-RelatedEye Disease Study,⁹⁰ a multicenter randomizedclinical trial involving more than 3600 patients, demonstrated that amongpatients with extensive intermediate drusen, at least 1 large druse, noncentralgeographic atrophy in 1 or both eyes, advanced AMD in 1 eye, or vision lossin 1 eye due to AMD, supplementation with antioxidant vitamins (ascorbic acid,500 mg/d; vitamin E, 400 IU/d; and beta carotene, 15 mg/d) and minerals (zincoxide, 80 mg/d; cupric oxide, 2 mg/d) reduces the risk of developing advancedAMD from 28% to 20% and the rate of at least moderate vision loss from 29%to 23%. Zinc is essential for the function of some antioxidant enzymes (eg,superoxide dismutase, catalase, and metallothionein) and is the most abundanttrace element in human eyes.⁹¹ As noted inthe "Bruch Membrane Thickness" subsection, zinc also is important for MMPactivity. Results of the Age-Related Eye Disease Study⁹⁰ indicatethat oxidative damage plays a role in the progression of AMD in its clinicallyevident intermediate and late stages and that disease progression can be alteredwith antioxidant supplementation. Earlier trials^68,92-96 ofzinc therapy and of dietary zinc intake gave conflicting results, possiblydue to small sample size, relatively short follow-up, and/or inadequate dosing.

Other studies^68,92-97 haveprovided data regarding antioxidant status and the risk of AMD, with conflictingresults in some cases. Complexities of study design (eg, the number of patientsstudied and reliance on historical information provided by patients), variabilityin diet-plasma correlation for micronutrients and antioxidants (eg, carotenoids),uncertain relationships between plasma levels of antioxidants and micronutrientsand their ocular tissue levels, and the possible importance of interactionsbetween various antioxidants and micronutrients might all underlie the variableresults reported in these studies. Other data supporting the hypothesis thatoxidative damage plays a role in AMD pathogenesis are as follows.

Epidemiologic Studies. Thus far, the most importantrisk factors for AMD (ie, those associated with at least a 2-fold increasedrisk) seem to be age, smoking, and race.^6,98-100 Regardingage, the prevalence of late AMD is approximately 0% at 50 years, 2% at 70years, and 6% at 80 years.¹⁰¹ The effect ofage on risk might indicate that oxidative damage must be gradual and cumulativefor AMD to develop. Also, it might be a sign that mitochondrial DNA damageplays a role in the pathogenesis.¹⁰² Smokingdepresses antioxidants (eg, decreases plasma vitamin C and carotenoids), induceshypoxia and reactive oxygen species, and alters choroidal blood flow.^103,104 Regarding the effect of race, whitepatients have a relatively higher risk of large drusen, pigmentary abnormalities,and exudative AMD complications compared with black patients.^44,105,106 Differencesin melanin content may underlie, in part, the racial differences in risk ofadvanced AMD. Melanin is a high-molecular-weight polymer arising from enzymaticoxidation of tyrosine and dihydroxyphenylalanine and is located in melanosomes,which are membrane-bound granules. In vitro experiments indicate that melaninreduces lipofuscin accumulation in RPE cells, possibly by interacting withtransition metals and scavenging radicals to function as an antioxidant.¹⁰⁷ The RPE melanin content in white and black patientsis similar, but black patients have substantially more choroidal melanin thanwhites.¹⁰⁸ Perhaps oxidative reactions in theBruch membrane (see the "Inflammation" subsection) or at the level of theRPE can be attenuated by choroidal melanin. Alternatively, the protectiveeffect of race may mean that the most important oxidative reactions leadingto AMD occur at the level of the choriocapillaris.¹⁰⁹ Thefact that choriocapillaris density decreases with AMD is consistent with butdoes not prove this hypothesis (see the "Abnormal ECM" and "Altered RPE-ChoriocapillarisBehavior" subsections).^23,89,110

Biochemical Studies. Antioxidants act by preventingthe formation of initiating radicals, binding metal ions, and removing damagedmolecules. Major antioxidants in the retina and RPE include water-solublemetabolites and enzymes (vitamin C [ascorbic acid], glutathione, catalase,glutathione peroxidase, and superoxide dismutase), lipid-soluble substances(vitamin E [α-tocopherol], retinoids [vitamin A derivatives], and carotenoids),and melanin.^75,78,79 Theantioxidant enzymes, for example, superoxide dismutase, catalase, and glutathioneperoxidase, constitute the primary defense against oxidative RPE damage.¹¹¹ Antioxidant molecules, for example, ascorbic acid(vitamin C), tocopherol (vitamin E), and carotenoids, support the enzymaticsystems.

The protective mechanisms against oxidative RPE damage seem to decreasewith aging, and, in some cases, these changes are greatest in AMD eyes (vsage-matched control eyes). For example, RPE catalase levels decrease withaging and with AMD.^73,112 Metallothioneinis an acute-phase reactant protein that scavenges hydroxyl radicals, and thereis an age-related decrease in submacular RPE metallothionein content.¹¹³ Plasma glutathione reductase is reduced substantiallyin patients with AMD.¹¹⁴ Frank and coworkers¹¹² found that heme oxygenase-1 and heme oxygenase-2immunoreactivity tended to decrease with increasing age, especially in RPElysosomes of neovascular AMD eyes. Oxidative stress probably causes pathologicup-regulation of lysosomal heme oxygenase-1 and possibly heme oxygenase-2.These investigators found that copper-zinc superoxide dismutase immunoreactivityincreases in the cytoplasm of submacular RPE in eyes with AMD and CNVs. ThePathologies Oculaires Liées à l'Age Study¹¹⁵ foundthat higher plasma levels of glutathione peroxidase were associated with asignificantly increased prevalence of late but not early AMD. Beatty and coworkers⁷⁵ pointed out that extracellular glutathione peroxidaseis believed to act as an extracellular antioxidant, which may be relevantfor oxidative reactions occurring in the Bruch membrane and the choriocapillaris(see the "Inflammation" subsection).

Carotenoids, especially lutein and zeaxanthin, compose the macular pigment.The primary direct antioxidant function of carotenoids is to scavenge singletoxygen, but they also quench the triplet state of photosensitizers and retardthe peroxidation of membrane phospholipids.^116,117 Factorsassociated with increased risk for AMD and increased risk for low macularpigment density include age, cigarette smoking, female sex, light iris color,and increasing lens density,^118-121 butnot all clinical studies confirm the association between low macular pigmentdensity and increased risk for AMD.¹²² Twopostmortem studies^123,124 revealeddecreased retinal lutein and zeaxanthin levels in AMD eyes vs control eyes.Increasing age and advanced AMD in the fellow eye have been associated witha relative absence of macular pigment.¹²⁵

Findings from in vitro and in vivo animal studies indicate that basallaminar deposit may form as a result of free radical–induced lipid peroxidationof RPE cell membranes with subsequent membrane blebbing and accumulation ofblebs as basal laminar deposit–like material in the sub-RPE space.¹²⁶ Advanced glycation end products occur at sites ofoxidant stress with hydroxyl radical formation. Advanced glycation end productsoccur in soft drusen, in basal laminar and basal linear deposits, and in thecell cytoplasm of RPE associated with CNVs.^88,127 WhenRPE cell lines are grown on a matrix modified by advanced glycation end products,they express genes (eg, transforming growth factor β₂) thatmight promote Bruch membrane thickening.¹²⁸ Advancedglycation end products induce increased expression of cytokines known to occurin CNVs.¹²⁷ Carboxymethyl lysine, a productof lipoprotein peroxidation or sequential oxidation and glycation, is presentin drusen and CNVs.^88,129 Onestudy⁸⁸ of drusen protein composition reportedoxidative protein modifications in TIMP-3 and vitronectin. Also, carboxyethylpyrrole protein adducts, which are uniquely generated from the oxidation ofdocosahexaenoate-containing lipids, were present and were much more abundantin drusen from AMD vs age-matched control donors. (As noted previously herein,docosahexaenoate lipid is abundant in photoreceptor outer segments.) The collocationof lipofuscin-induced autofluorescence and drusen suggests an etiologic relationshipbetween the two, but this collocation has not been observed in all studies.^130,131 Genetic defects (eg, in antioxidantenzymes), dietary or uptake deficiencies in antioxidants, or exposure to noxiousagents (eg, cigarette smoking) could enhance oxidative RPE damage during lifeand predispose to AMD and other signs of aging.

Anatomic studies^132-135 providedinitial evidence for the role of inflammation in CNV formation in AMD. Subsequently,molecular evidence for the role of inflammation in AMD pathogenesis has beendeveloped and summarized by Hageman,¹³⁶ Johnson,¹³⁷ and Anderson¹³⁸ andtheir coworkers. Protein components of drusen include immunoglobulin and componentsof the complement pathway associated with immune complex deposition (eg, C5b-9complex), molecules involved in the acute-phase response to inflammation (eg,amyloid P component and α₁-antitrypsin), proteins that modulatethe immune response (eg, vitronectin, clusterin, apolipoprotein E, membranecofactor protein, and complement receptor 1), major histocompatibility complexclass II antigens, and HLA-DR and cluster differentiation antigens.^139-142 Cellularcomponents of drusen include RPE blebs, lipofuscin, and melanin, as well aschoroidal dendritic cells.^136,143-146 Hagemanand coworkers¹³⁶ postulated that choroidaldendritic cells are activated and recruited by injured RPE (eg, via monocytechemotactic protein) and oxidized proteins and lipids in the Bruch membrane.A similar process occurs in atherosclerosis. The RPE cells respond to controldendritic cell activation by secreting proteins that modulate the immune response,including vitronectin, apolipoprotein E, and membrane cofactor protein.¹³⁷ Johnson and coworkers^137,147 pointedout that the cytoplasmic accumulation of vitronectin, apolipoprotein E, andother drusen-associated molecules suggests that the cells are subjected toa chronic sublethal complement attack. These researchers recognized that complementattack can result in the elimination of surface-associated membrane attackcomplexes (by shedding or endocytosis of cell membrane) and in the formationof extracellular deposits of immune complexes and complement intermediates.Penfold and coworkers¹⁴⁸ reported an increasein major histocompatibility complex class II immunoreactivity on retinal vascularelements and morphologic changes in microglia in eyes with incipient AMD.These immunologic changes seemed to be related to early pathological changesin RPE pigmentation and drusen formation. Evidence of inflammatory cell involvementin the later stages of AMD includes the presence of multinucleated giant cellsand leukocytes in the choroid of AMD eyes^149-151 andin excised CNVs.^152,153 Macrophagesand foreign body giant cells near the Bruch membrane become more common whenbasal linear deposit is present.¹⁵¹ Activatedmacrophages and other inflammatory cells secrete enzymes that can damage cellsand degrade the Bruch membrane, and, by releasing cytokines, inflammatorycells might foster CNV growth into the sub-RPE space.^154,155 Thus,in AMD eyes, breaks in the Bruch membrane probably are the result and notthe cause of CNVs.¹⁵⁶ In some systems, ECMdegradation is associated with free radical release.^157,158

Poorly degradable RPE debris and Bruch membrane components (eg, wide-spacedcollagen) might stimulate chronic inflammation.^{27,46,150,159} Hagemanand coworkers¹³⁶ suggested that activationof choroidal dendritic cells might initiate an autoimmune response to retinaland/or RPE antigens or to neoantigens created within the Bruch membrane. Despitethe RPE and retina being immune-privileged tissues,¹⁶⁰ antiretinaland anti–RPE antibodies have been detected in the serum of patientswith AMD.^136,161,162 Johnsonand coworkers¹³⁷ pointed out that complementactivation and associated inflammatory events occur in diseases exhibitingcellular degeneration and accumulation of abnormal tissue deposits, for example,atherosclerosis and Alzheimer disease.¹⁶³ Inthese diseases, damaged cells and highly insoluble protein deposits and extracellulardebris activate the classical and alternative complement pathways, resultingin chronic direct and bystander cellular damage with attendant cell surfaceblebbing, endocytosis, and up-regulation of defense proteins. The Alzheimeramyloid β peptide co-localizes with activated complement components ina substructural vesicular component with drusen.¹⁶⁴

Intravitreal corticosteroids reduce the incidence of laser-induced CNVsin primates, possibly by altering inflammatory cell activity and/or numbersin the choroid.¹⁶⁵ Other potential mechanismsinclude reduction of vascular endothelial growth factor (VEGF) expression(see the "Biochemical Features of CNV Growth" subsection) and down-regulationof intercellular adhesion molecule 1,^166,167 whichis constitutively expressed on RPE and choroidal endothelial cells and mediatesleukocyte adhesion and diapedesis during inflammation.^168,169

The RPE deposits cytoplasmic material into the Bruch membrane throughoutlife, possibly to eliminate cytoplasmic debris or as a response to chronicinflammation (see the "Inflammation" subsection).^143,170-172 Histologically,AMD eyes exhibit abnormal extracellular material in 2 locations: (1) betweenthe RPE plasmalemma and the RPE basement membrane and (2) external to theRPE basement membrane within the collagenous layers of the Bruch membrane.The former material is termed basal laminar deposit, andthe latter material is termed basal linear deposit.⁴⁴ Although basal laminar deposit persists in areasof geographic atrophy, basal linear deposit disappears, which is consistentwith the notion that basal linear deposit arises mostly from the RPE-photoreceptorcomplex.¹⁷³ Basal linear deposit may be morespecific to AMD than basal laminar deposit.⁴⁸ Softdrusen can represent focal accentuations of basal linear deposit in the presenceor absence of diffuse basal linear deposit–associated thickening ofthe inner aspects of the Bruch membrane.^44,174 Softdrusen can also represent a localized accumulation of basal laminar depositin an eye with diffuse basal laminar deposit.¹⁷⁴ Thus,the abnormal ECM of AMD eyes includes basal laminar deposit, basal lineardeposit, and their clinically evident manifestation, soft drusen.

Drusen represent the earliest clinical finding in AMD. Drusen compositionand origin have been analyzed extensively.^{34,56,127,136,138,140,175-177} Small(ie, <63-µm-diameter) drusen generally do not signify the presenceof AMD.^5,26,44,178 Excessivenumbers of small hard drusen, however, can predispose to RPE atrophy at arelatively young age.¹⁷⁹ Soft drusen are usuallypale yellow and large (≥63 µm in diameter), with poorly demarcatedboundaries. Many different molecules have been identified in drusen, includingglycoconjugates¹⁸⁰ containing mannose, sialicacid, N-acetylglucosamine, and β-galactose (Table 2). Abnormal constituents of theECM probably underlie the increased blue-green autofluorescence of the Bruchmembrane in AMD eyes.¹⁸¹

Most of the molecular constituents of drusen are synthesized by RPE,neural retina, or choroidal cells, but some are derived from extraocular sources.^{56,127,136,176,177,182} Severalinvestigators^183,184 have notedthat drusen tend to be distributed near the collecting venules of choriocapillarislobules, which has led to the hypothesis that drusen are derived from thechoroidal vasculature. An alternative explanation is that RPE cell susceptibilityto metabolic derangement depends, to some degree, on the location of a givencell with respect to the underlying choriocapillaris lobule.

A variety of drusen constituents (eg, vitronectin, apolipoproteins Band E, complement, and lipid) are present in atherosclerotic plaques, whichmay reflect the association of some atherosclerosis risk factors with thedevelopment of AMD. Amyloid P component, C5, and α₁-antitrypsinare acute-phase reactants (ie, up-regulated expression in response to inflammation),and vitronectin, C5, and apolipoprotein E have roles in mediating immune responses.These findings have led to the suggestion that immune complex–mediateddamage to RPE cells plays a role in the initiating events of drusen formation,as noted herein. It may be that terminal complement activation promotes drusenbreakdown by enzymatic digestion and phagocytosis.¹⁴⁰ Animmune response directed against RPE-derived antigens might be the triggerfor drusen formation.

Although MMPs and TIMPs are present in plasma,¹⁸⁵ immunohistochemicalstudies of MMP and TIMP indicate that at least some MMPs and TIMPs in theBruch membrane are derived from the RPE (see the "Bruch Membrane Thickness"subsection). Changes in MMPs and their inhibitors indicate that in AMD, RPEdysfunction could result in abnormal MMP activity, which could contributeto the exaggerated development of an abnormal ECM (compared with age-matchedcontrols). For example, TIMP-3 is present in drusen,^34,186 andTIMP-3 levels seem to be increased in drusen and in the Bruch membrane ofAMD eyes.³⁴ Binding of TIMP-3 to advanced glycationend products, which are present in the Bruch membrane and drusen (see the"Role of the ECM in CNV Growth" subsection), may lead to TIMP-3 accumulationin AMD eyes.^33,187 Leu and coworkers¹⁸⁶ noted that MMP immunoreactivity was present onlyon the surface of drusen. In situ zymography demonstrated that metal ion–dependentgelatinase activity was absent in drusen cores. Leu and coworkers¹⁸⁶ suggested that the lack of proteolysis in drusencores might contribute to drusen formation and AMD progression, perhaps inthe same way that Sorsby fundus dystrophy mutations, which do not result inloss of TIMP-3 function, foster accumulation of an abnormal ECM.^33,186

In patients with AMD, delayed choroidal perfusion (as visualized withfluorescein and indocyanine green angiography) and psychophysical retinalfunctional abnormalities may result from the diffusion barrier created bya thickened, lipid-laden Bruch membrane.^58,188-190

The accumulation of extracellular debris alters Bruch membrane composition(ie, increased lipid and protein content) and permeability (eg, decreasedpermeability to water-soluble constituents in plasma, decreased amino acidtransport, and possibly decreased bulk flow of extruded RPE-derived cytoplasmicdebris across the Bruch membrane).^43,53,59-61 Thesechanges may lead to impaired diffusion of waste products from and of hormonesand nutrients to the RPE, including oxygen and vitamin A.¹³³ Inresponse to this metabolic distress, the RPE probably produces substancesthat stimulate CNV growth. Several investigators^191-193 haveshown that RPE cells associated with CNVs produce VEGF and basic fibroblastgrowth factor, which may act synergistically to stimulate new blood vesselgrowth. Macrophages and foreign body giant cells, possibly recruited by activatedchoroidal dendritic cells, may digest the Bruch membrane and be a source ofcytokines that also stimulate CNV growth. Sarks and colleagues¹⁹⁴ notedthat activated choroidal capillaries were observed only beneath thinned regionsof the Bruch membrane. Sarks observed that Bruch membrane erosion often commencesbeneath small hard drusen and posited that this phenomenon occurs becauseangiogenic factors diffuse most readily to the choriocapillaris in these locirather than through nearby areas containing substantial membranous debris.¹⁹⁴ Choroidal capillaries adjacent to CNVs are oftennarrowed or absent, which may be a manifestation of RPE dysfunction.^{66,89,195,196} Thisfact may underlie the clinical finding that pulsatile ocular blood flow islower in eyes with CNVs compared with contralateral eyes with drusen.¹⁹⁷

Biochemical Features of CNV Growth. The RPEmay constitutively control angiogenesis beneath the retina.¹⁹⁸ TheRPE, for example, produces VEGF in vivo under physiologic conditions.¹⁹⁹ VEGF is secreted as a homodimeric protein that isexpressed in ischemic retina and stimulates endothelial cell proliferationin blood vessels. The RPE-derived VEGF may maintain the fenestrated choriocapillarisendothelium. Much evidence implicates VEGF in CNV formation. High concentrationsof VEGF and VEGF receptors are in CNVs, surrounding tissue, and RPE cells.^191,193,200 Levels of VEGFare increased in cadaver AMD eyes,²⁰⁰ in thevitreous of patients with AMD,²⁰¹ and in theplasma of patients with AMD.²⁰² Also, VEGFis present in fibroblastic cells and transdifferentiated RPE of surgicallyremoved CNVs.¹⁹¹ Laser-induced CNVs (in ratsand monkeys) are associated with increased VEGF messenger RNA (mRNA) in theRPE, choroidal vascular endothelial cells, and fibroblasts. Intravitreal anti-VEGFantibody fragment (rhuFab VEGF) prevents laser-induced CNVs in monkeys anddecreases leakage from already-formed CNVs.²⁰³ PKC412, an inhibitor of protein kinase C and the kinases of VEGF and platelet-derivedgrowth factor receptors, prevents laser-induced CNVs in mice.²⁰⁴

Increased VEGF expression seems to be sufficient for CNV formation.Schwesinger and coworkers²⁰⁵ showed that transgenicmice with RPE cells that overexpress VEGF are associated with increased VEGFin the RPE, Bruch membrane, and choroid; increased leukostasis in the choroidalvasculature, probably from up-regulation of intercellular adhesion molecule-1;and intrachoroidal new vessels that do not penetrate the intact Bruch membrane.This result suggests that the presence of increased RPE-derived VEGF mustbe coupled with some additional factor(s) to develop "typical" sub-RPE and/orsubretinal CNVs. The need for RPE–Bruch membrane damage may explainwhy rats undergoing subretinal injection of adenovirus vector that transfectsRPE exhibit CNVs.²⁰⁶

Retinal pigment epithelium cells also produce pigment epithelial–derivedfactor (PEDF), a 50-kDa protein.²⁰⁷ In additionto neuroprotective effects, PEDF potently inhibits angiogenesis.²⁰⁸ Oxidativestress may alter the balance between RPE VEGF and PEDF production.²⁰⁹ Increased RPE PEDF production inhibits laser-inducedCNV growth and induces regression of established CNV in a murine model.^210,211

Tie-1 and Tie-2 are receptor tyrosine kinases that play a role in thelater stages of angiogenesis.²¹² Angiopoietinsare approximately 75-kDa secreted proteins that target endothelial cell–specificTie-2 receptors and promote neovascularization in vascular beds, includingthe retina. Stimulated by angiopoietin-1 (Ang-1), Tie-2 plays a major rolein recruiting and sustaining periendothelial support cells (eg, pericytes),resulting in the formation of multicellular vascular structures from simpleendothelial tubes.^213,214 Angiopoietin-2(Ang-2) blocks these functions and, in so doing, may allow vascular remodelingand angiogenesis via stimulatory cytokines, such as VEGF (eg, by reducingendothelial-matrix contact or by dissociating pericytes from endothelial cells).²¹³ Thus, Ang-1 promotes maturation and stabilizationof vessels, and Ang-2 might allow endothelial cells to respond to angiogenicsignals.^213,214 Cultured RPE cellsexpress Ang-1 and Ang-2 mRNA, and VEGF up-regulates RPE Ang-1 mRNA and Ang-1protein secretion.²¹⁵ In one study,²¹⁶ Ang-1, Ang-2, and Tie-2 immunoreactivity were presentin CNVs from AMD eyes. Angiopoietin-1, Ang-2, and VEGF localization were similar,including localization in RPE and vascular cells. Angiopoietin-2 and VEGFimmunoreactivity was abundant in highly vascularized regions of the CNVs.Tie-2 immunoreactivity was present in vascular structures and RPE cells. Angiopoietin-1probably modulates the effect of VEGF on endothelial cells during CNV formation.Retinal pigment epithelium–derived Ang-1 may modulate interactions betweenendothelial cells and leukocytes during choroidal angiogenesis.

Hypoxia up-regulates Ang-2 mRNA in bovine retinal capillary endothelialcells.²¹⁷ Hypoxia also up-regulates VEGF levels(ie, up-regulates VEGF mRNA transcription and increases mRNA stabilization).It is not proved that the documented abnormalities in choroidal blood flowin AMD eyes are sufficient to induce this hypoxia response in the RPE-choroid.Also, it is not proved that the documented thickening, lipidization, and proteincross-linking of the Bruch membrane in AMD alter oxygen diffusion to the RPEphotoreceptors. How might oxidative damage and hypoxia play a role in AMDpathogenesis? Perhaps initial oxidative damage leads to excessive formationof an abnormal ECM. Thickened Bruch membrane, combined with factors such assmoking, might then create a relatively hypoxic environment. Relatively minorchanges in the diffusion properties of the Bruch membrane or in choroidalblood flow might have seemingly disproportionate effects on the RPE and photoreceptorssince the photoreceptors usually consume 90% to 100% of the oxygen deliveredby the choriocapillaris.²¹⁸ (In the dark-adaptedmacaque monkey, the oxygen tension near the level of the inner segments isapproximately 8 mm Hg vs approximately 50 to 80 mm Hg in the choroid.) Hypoxiacould then result in RPE death and geographic atrophy or in stimulation ofCNV growth by hypoxic RPE.

Role of the ECM in CNV Growth. Sarks et al¹⁹⁴ and Campochiaro and coworkers²¹⁹ suggestedthat abnormalities of the RPE ECM may promote a proangiogenic phenotype thatfosters CNV growth; CNV growth probably is affected by the nature and quantityof extracellular debris in the sub-RPE space. The risk of CNVs in AMD, forexample, increases with increasing number, size, and confluence of drusen.Vitronectin, fibronectin, and advanced glycation end products are molecularconstituents of drusen and stimulate production of angiogenic factors in modelsystems.^220,221 Peroxidized lipids,which accumulate in the Bruch membrane, not only alter Bruch membrane hydraulicconductivity but also stimulate the production of substances that promoteneovascularization.⁵⁶ Also, ECM molecules canstimulate or inhibit angiogenesis by binding to integrins or by altering integrinexpression on endothelial cells (Table 3).^{204,209,219,222,223}

Matrix metalloproteinases and urokinase plasminogen activator breakdown the ECM during angiogenesis.^224,225 Matrixmetalloproteinase-2 and MMP-9 mRNA, for example, are present in excised CNVspecimens,²²⁶ and MMP-2 mRNA is increased inlaser-induced CNVs in rats.²²⁷ Degradationof ECM presumably releases and/or activates proangiogenic factors. Proangiogenicfactors stimulate proteolytic activity, migration, proliferation, and tubeformation in endothelial cells.

Biological Basis of Geographic Atrophy. Choriocapillarisdensity decreases with aging and with AMD. The average choroidal blood flowis lower in patients with dry AMD vs age-matched controls.^228,229 Anarea dilution analysis technique applied to indocyanine green angiographydemonstrated delayed and heterogeneous choroidal filling in nonneovascularAMD eyes compared with age-matched control eyes.²³⁰ InAMD, the usual pattern of sinusoidal capillary lobules (ie, a central arteriolefeeding a sinusoid that drains into peripheral venules) is replaced by a tubularcapillary network, which has a lower surface area–volume ratio.¹¹⁰ This change might be due to primary damage to thechoriocapillaris endothelium (eg, oxidative damage mediated by protoporphyrins);might follow the loss of RPE cells (eg, secondary to chronic oxidative damage),with attendant loss of VEGF and other trophic factors⁶⁶;or might arise from a combination of these processes. Photoreceptor deathfollows RPE cell loss.^132,231

The biological basis of geographic atrophy in AMD has been reviewedby Sunness.²³² The presence of drusen measuring250 µm or greater and pigmentary abnormalities are risk factors forthe development of geographic atrophy. Increased fundus autofluorescence precedesthe development and enlargement of geographic atrophy in AMD.²³³ Thedominant fluorophores of fundus autofluorescence are part of RPE lipofuscingranules. Thus, excessive RPE lipofuscin accumulation may play a criticalrole in geographic atrophy pathogenesis in AMD. Histologically, lipofuscin-ladenRPE cells are present at the junction of atrophic and normal retina in AMDeyes. The RPE appears increasingly abnormal near the area of atrophy.^44,179 Clinical studies²³³ indicatethat the area of increased autofluorescence is larger than would be predictedfrom histologic studies, but no direct clinicopathologic correlation has beenmade yet. The loss of RPE-derived VEGF might result in the choriocapillarisatrophy seen with atrophic AMD.⁸⁹

Geographic atrophy tends to develop near the fovea, but it tends tospare the foveal center until the later stages of the disease.^231,234,235 Oneexplanation for this observation is as follows.^21,231,236 Thehighest turnover of outer segments involves rods just outside the fovea, whichparallels the distribution of lipofuscin in the RPE. Each RPE cell contactsapproximately 45 photoreceptor cells, and each rod outer segment is fullyphagocytosed and replaced approximately every 10 days. The RPE continuouslydischarges cytoplasmic material into the Bruch membrane,²⁷ whichcould lead to pathological changes, primarily in the subjacent Bruch membrane.Atrophy of the RPE may be a response to decreased nutrients/increasing metabolicabnormalities in areas of excessive accumulation of extracellular debris.Subfoveolar RPE is spared from atrophy the longest, perhaps by macular pigment,²³⁷ the high cone density in the foveola, and possiblyother factors. It may be that the subfoveolar RPE is the longest-lived sourceof neovascular signal(s) in this metabolically distressed region, thus accountingfor the tendency of CNVs to grow toward the foveola initially and after laserphotocoagulation. Choroidal new vessels do not seem to arise from within areasof geographic atrophy but instead tend to arise under adjacent areas in whichthe RPE-retina seems relatively preserved or under the fovea, if spared.²³⁵

Among patients with CNVs in one eye and geographic atrophy in the felloweye, the cumulative incidence of CNVs in the eye with geographic atrophy is30% to 50% at approximately 5-years' follow-up.²³⁷ Resultsof histopathologic studies^26,44 indicatethat CNVs are present in approximately one third of cases with geographicatrophy. Thus, it seems unlikely that patients with AMD-associated geographicatrophy and AMD-associated CNVs have 2 different diseases.

Findings from genetics studies (reviewed by Yates and Moore²³⁸) indicate that there is high concordance for AMDamong monozygotic twins and relatively lower concordance among dizygotic twins.Age-related macular degeneration is more likely in first-degree relativesthan in age-matched controls.²³⁹ One study²⁴⁰ of monozygotic twins found a genetic effect forthe phenotypes of age-related maculopathy (ie, the early stages of AMD), softdrusen, pigmentary changes, and 20 or more hard drusen. The inheritabilityof age-related maculopathy was estimated to be 45%. During the next few years,molecular biology studies probably will identify mutations in specific genesthat alter the risk of developing AMD.²⁴¹ Atthis time, it seems likely that AMD is a polygenic disorder with multiplegenes conferring susceptibility to and resistance from the disease. Characterizationof the genetic defects underlying the diseases to which we refer currentlyas AMD may provide an opportunity to identify subtypes of AMD with differentdisease-causing molecular defects. This approach to nosology may permit betterdesign of clinical trials of therapy, which in turn would enable physiciansto provide the proper treatment (especially prophylaxis) at the proper timefor any given patient.

Genetic mutations have been identified as the causes of diseases thatresemble AMD. Kuntz and coworkers²⁴² observedthat degenerations associated with lipid/mineral deposits in the Bruch membraneare often autosomal dominant and that patients are often asymptomatic untiladulthood. Genetic factors probably play a role in the development of geographicatrophy in AMD. Zermatt macular dystrophy, for example, is associated witha dominant mutation of the RDS/peripherin gene andis associated with atrophy.²⁴³ A mutation inepidermal growth factor–containing fibrillin-like ECM protein-1 (EFEMP1)causes Malattia Leventinese and Doyne honeycomb retinal dystrophy, which areassociated with drusen formation.²⁴⁴ EFEMP1is expressed in the RPE and retina and encodes a protein homologous to a familyof ECM glycoproteins known as fibulins. ABCR genemutations have been associated with an increased risk of atrophic AMD, butfindings from some studies^245,246 indicatethat the detected mutations may simply be polymorphisms (ie, mutations thatcan occur in healthy individuals and do not confer an increased risk) amongpatients with a common disease. ABCR, or rim protein, is a transmembrane proteinthat may be involved in retinoid transport. As noted in the "Bruch MembraneThickness" subsection, mutations in TIMP-3 cause Sorsby fundus dystrophy,which is associated accumulation of abnormal extracellular material in theBruch membrane, patchy choroidal filling on fluorescein angiography, and CNVs.Despite the similarities between Sorsby fundus dystrophy and AMD, there aredifferences (eg, younger age of CNV development in the former), and TIMP-3mutations do not seem to cause AMD.^247,248 Althoughlysosomal storage diseases due to single gene defects in metabolism are characterizedby accumulations of intracellular material in the RPE, their resemblance toAMD is modest.²⁴⁹ Cai and coworkers⁷⁹ noted that Cockayne syndrome may be a more relevantdisease model. In this condition, nucleotide excision repair and transcriptionalrepair are deficient, which permits cumulative damage to nuclear and mitochondrialgenomes throughout life and premature aging.²⁵⁰ Inanalogy with mitochondrial myopathies, Cai and coworkers⁷⁹ suggestedthat AMD probably results from progressive damage to the retina and RPE throughoutlife. Functional capacity (eg, against oxidative damage) in healthy individualsexceeds the threshold below which disease becomes apparent clinically. Geneticbackground interacts with exposure to environmental risk and protective factorsto determine age at disease onset for a given individual. Genetics might affectthe susceptibility to develop AMD in the following way.²⁵¹ Cellularproduction of ECM is genetically controlled. Epigenetic factors can alterthe ECM, for example, formation of advanced glycation end products. The matrixand matrix degradation products regulate cell phenotype via membrane receptors.The consequences of these interactions can cause disease and/or aging. Forexample, advanced glycation end products can trigger free radical releaseand seem to play a role in the accumulation of abnormal extracellular materialin Alzheimer disease.²⁵²

Most AMD cases may not be caused by a single gene defect, even if genesplay a major role in determining susceptibility to the disease (eg, by affectingmelanin content, antioxidant enzyme activity, and/or MMP activity). Nonetheless,an important benefit of identifying a gene that causes AMD (even if it wereto account only for a small fraction of the total patient population) or acondition that very strongly resembles AMD or just a particular aspect ofthe disease is that one might then construct a biochemical pathway "framework"in which other causes of the disease might be understood and various treatmentstrategies might be developed.²⁵³ The evidencesummarized in this review may provide some insight into the nature of thebiochemical pathways involved in AMD pathogenesis and a context in which identifiedmutations can be studied. The evolution of technologies such as serial analysisof gene expression and microarray analysis may accelerate the identificationof genes conferring susceptibility to or protection against AMD and fosterthe development of complex animal models exhibiting more than 1 gene defect.²⁵⁴

Chronic diseases often require a stepwise approach to treatment, andprogressive steps usually are associated with increasing degrees of risk tothe patient. Generally, the initial approach is prophylaxis, followed by medicaltherapy, and then surgical therapy. Examples that illustrate this sequenceinclude diabetes mellitus (eg, weight control → oral hypoglycemic agents→ insulin therapy → pancreatic islet cell transplantation) and atherosclerosis(eg, diet restriction + exercise → cholesterol-lowering agents →coronary artery stenting and/or bypass surgery). Age-related macular degenerationis a chronic disease; thus, the goal of AMD research should be to developtreatments for the early and later stages of the disease, as patients willseek care at different stages of the disease and will vary in their responseto therapy.

Currently, most treatments for AMD benefit patients with advanced stagesof the disease (eg, laser photocoagulation, photodynamic therapy, and surgeryfor CNVs). Lanchoney and coworkers²⁵⁵ showedthat the addition of a 25% effective bilateral preventive treatment to theconventional laser photocoagulation treatment regimen for CNVs would reducethe rate of legal blindness in the population with bilateral soft drusen relativeto current laser treatment by approximately 40% (from 2.24% to 1.34%). Preventivetreatment given to the fellow eye after CNVs develop would have substantiallyless impact (approximately 20% reduction). Thus, treatments for the earlierstages of AMD, even if only modestly effective, can have a great impact onthe incidence of AMD-induced blindness.

Better knowledge of AMD pathogenesis will permit the design of effectivetherapy for earlier stages of the disease. The first proven "early" treatmentfor AMD is oral therapy with antioxidant vitamins and minerals. Other approachesalso may be effective. Dimethyl fumarate, for example, is enriched in apples,increases glutathione, and protects RPE cells from peroxide-induced damagein vitro.²⁵⁶ Other more potent inducers ofglutathione synthesis, such as oltipraz or sulforaphane, have been testedin preclinical trials for cancer prevention.²⁵⁷ Ahirand coworkers²⁵⁸ showed that in vitro "transplantation"of cultured RPE cells, which express active MMP-2 and MMP-9, results in improvedBruch membrane hydraulic conductivity. Lipofuscin accumulation in RPE cellscan be reduced by treatment with lutein, zeaxanthin, lycopene, or α-tocopherol²⁵⁹ or reversed by centrophenoxine treatment.²⁶⁰ The role of chronic inflammation in AMD pathogenesishas led to the consideration of anti-inflammatory therapy as treatment forthe early stages of the disease.^136,138

Better knowledge of the biological changes underlying AMD will alsofoster the development of sight-restoring treatments for the late stages ofAMD. For example, anti-inflammatory therapy, anti-VEGF therapy, PEDF therapy,anti–Tie-2 therapy (either via ligation of free Ang-1 or via high-levelexpression of Ang-2, especially in the face of VEGF inhibition), or controlof TIMP and MMP expression may permit control of CNV growth without retinal,RPE, or choriocapillary damage.^210,261-266 Moreprecise characterization of AMD-induced Bruch membrane ECM abnormalities probablywill permit the design of more effective cellular transplantation surgery,which might serve as a treatment for exudative and atrophic AMD.^267-269 Neurotrophicagents, which promote neuron survival, might be useful for preserving visionin patients with atrophic and exudative manifestations of AMD.²⁷⁰ Genetherapy with PEDF might be an example of combined therapy, in which the treatmentinhibits CNV growth and promotes photoreceptor survival via a neurotrophiceffect (see the "Biochemical Features of CNV Growth" subsection).

Five general concepts relevant to the cell biology of AMD have beendescribed (Figure 2). First, AMDinvolves aging changes plus additional pathological events. Second, in agingand AMD, oxidative stress causes RPE and, possibly, choriocapillaris injury.Third, in AMD (and perhaps in aging), RPE and, possibly, choriocapillarisinjury results in a chronic inflammatory response in the Bruch membrane andthe choroid. Fourth, in AMD, RPE and, possibly, choriocapillaris injury andinflammation lead to formation of an abnormal ECM. This abnormal ECM causesaltered diffusion of nutrients to the retina and RPE, which may precipitatefurther RPE and retinal damage. Fifth, the abnormal ECM results in alteredRPE-choriocapillaris behavior, leading ultimately to atrophy of the retina,RPE, and choriocapillaris and/or to CNV growth. In this pathogenetic sequenceof events, environment and genetics can alter any given patient's susceptibilityto the disease. Manipulation of environmental variables (eg, antioxidant levels)provides an opportunity for early therapeutic intervention. Gene and/or cellulartherapy provide an opportunity for later, sight-restoring treatment. Implicitin this characterization of AMD pathogenesis is the concept that there islinear progression from one stage of the disease to the next. This assumptionmay be incorrect, and different biochemical pathways leading to geographicatrophy and/or CNVs may operate simultaneously. Additional experimentationwith in vitro and in vivo models will prove or refute the concept of linearprogression and will establish the identities of the various pathways thatlead to CNVs and geographic atrophy.

Better knowledge of AMD cell biology will lead to better treatmentsfor AMD at all stages of the disease, as results from the Age-Related EyeDisease Study⁹⁰ and the Verteporfin in PhotodynamicTherapy study⁴ imply. Many unanswered questionsregarding AMD pathogenesis remain. For example, if Bruch membrane permeabilitychange plays in the evolution of AMD, is it reversible? If oxidative damageprecipitates the development of geographic atrophy and CNV growth, what additionalfactors determine the development of atrophy vs CNV in any given patient orlocation within a given eye? For any given patient, can a risk profile anda "prophylactic treatment" plan be established based on genotype analysis?The development of transgenic animal models of AMD and better in vivo RPE–Bruchmembrane imaging modalities will accelerate progress in answering these questions.It seems that multiple genes confer susceptibility to and resistance fromAMD. Thus, animal models exhibiting multiple genetic changes may be neededto consistently reproduce all aspects of AMD. New techniques, such as serialanalysis of gene expression, may be critical for the identification of thesegenes. Until such complex animal models are developed, there probably willbe an ongoing need for simpler animal models and for in vitro models of specificaspects of AMD so that the development of effective therapies for differentstages of AMD can continue.

Corresponding author and reprints: Marco A. Zarbin, MD, PhD, TheInstitute of Ophthalmology and Visual Science, New Jersey Medical School,90 Bergen St, Suite 6100, Newark, NJ 07103-2499 (e-mail: zarbin@njmsa.umdnj.edu).

Submitted for publication February 6, 2002; final revision receivedJuly 15, 2003; accepted August 4, 2003.

This study was supported in part by Research to Prevent Blindness Inc,New York, NY, the New Jersey Lions Eye Research Foundation, Newark, and theEye Institute of New Jersey, Newark.