Age-Related Macular Degeneration, a Mathematically Tractable Disease

Abstract

A progression sequence for age-related macular degeneration onset may be determinable with consensus neuroanatomical nomenclature augmented by drusen biology and eye-tracked clinical imaging. This narrative review proposes to supplement the Early Treatment of Diabetic Retinopathy Study (sETDRS) grid with a ring to capture high rod densities. Published photoreceptor and retinal pigment epithelium (RPE) densities in flat mounted aged-normal donor eyes were recomputed for sETDRS rings including near-periphery rich in rods and cumulatively for circular fovea-centered regions. Literature was reviewed for tissue-level studies of aging outer retina, population-level epidemiology studies regionally assessing risk, vision studies regionally assessing rod-mediated dark adaptation (RMDA), and impact of atrophy on photopic visual acuity. The 3 mm-diameter xanthophyll-rich macula lutea is rod-dominant and loses rods in aging whereas cone and RPE numbers are relatively stable. Across layers, the largest aging effects are accumulation of lipids prominent in drusen, loss of choriocapillary coverage of Bruch's membrane, and loss of rods. Epidemiology shows maximal risk for drusen-related progression in the central subfield with only one third of this risk level in the inner ring. RMDA studies report greatest slowing at the perimeter of this high-risk area. Vision declines precipitously when the cone-rich central subfield is invaded by geographic atrophy. Lifelong sustenance of foveal cone vision within the macula lutea leads to vulnerability in late adulthood that especially impacts rods at its perimeter. Adherence to an sETDRS grid and outer retinal cell populations within it will help dissect mechanisms, prioritize research, and assist in selecting patients for emerging treatments.

Photoreceptors and retinal pigment epithelium (RPE) together form a sensor chip for the brain, like that found in a smart phone camera. Retinal topography (i.e., the distribution of cells across the retina) is honed by evolution over millennia so that each species’ vision, including that of humans, is matched to its habitat.^1,2 The evolutionary and developmental biology of human fovea created a singularity that serves as the origin of a geometric coordinate system. Modern clinical imaging anchored on eye-tracked optical coherence tomography (OCT) adds fourth dimension (time, t) to anatomical x,y,z. 

This narrative review posits that in the transition from aging, age-related macular degeneration (AMD) is an orderly and mathematically tractable disease because of the steep gradients of photoreceptor cell density (cells/ mm²) centered on the fovea in a nearly radially-symmetric manner. This arrangement creates biologic effects that are large, regionally specific, and accessible in exquisite cellular and subcellular detail over time in the living eye. Effects of the human neural configuration may be apparent even in geographic atrophy, end-stage of the underlying degeneration, in the form of foveal sparing. Currently approved AMD treatments target advanced disease. Yet the biology and neural geometry of AMD let us contemplate precision prevention at a population level. This route is inspired by decades-long success in abating atherosclerotic cardiovascular disease,^3–6 which shares with AMD the molecular commonality of lipoprotein-propelled pro-inflammatory progression in a vessel wall. In the eye, Bruch's membrane (BrM) represents the vessel wall, and lipoproteins are produced locally. Solving the four-dimensional puzzle can be accelerated with standardized anatomic terminology. We propose to update a clinical ruler, the Early Treatment of Diabetic Retinopathy Study (ETDRS) grid. 

This review explores the topographic relationship between foveal cone vision and the formation of high-risk drusen. An equally strong relationship exists between the distribution of rods and subretinal drusenoid deposit (SDD).^7–9 This stereotypic extracellular material between the photoreceptors and RPE was recently named the third risk factor of AMD progression.¹⁰ SDD and drusen have been proposed as two parts of one system for lipid transfer serving rods and cones in a cell type- and regionally-specific manner.^7,11 Although this important topic is beyond the current scope, everything suggested herein is compatible with similar reasoning for SDD. Indeed, SDD research would be furthered by application of an ETDRS grid supplemented by an additional ring (sETDRS) to multimodal imaging anchored by OCT. Finally, the parallel course of age-related rod loss and choriocapillaris degeneration explored herein was also noted in an independent analysis of cell populations and transport mechanisms¹² that interested readers may wish to consult. 

Defining the Macula Lutea

Ophthalmology textbooks define the macula lutea, or yellow spot, as a retinal region containing xanthophyll carotenoid pigments specific to human retina. A medical dictionary defines a macula as a spot that is clearly differentiated from its background, a definition that depends on detection technology. Duke-Elder described the macula lutea in a fresh cadaver eye as “visible to the unaided eye” in an elliptical region 3 mm in diameter centered on the fovea.¹³ Furthermore, “outside this region … a faint coloration can be determined {with special optical apparatus} within a wider circle reaching almost to the temporal margin of the disc and the retinal area supplied with large vessels.” Accordingly, the light background is described as 5 to 6 mm in diameter. A modern clinical imaging method for xanthophylls, two-wavelength autofluorescence, reveals a distribution consistent with this definition¹⁴ with variance between individuals in part because of foveal shape.^15,16 

For this review, we operationally define the macula lutea as the retinal region vulnerable to AMD, a 3 mm-diameter circle centered on the fovea; this supersedes previous definitions by Curcio et al.¹⁷ and Quinn et al.¹⁸ We will also use the name macula lutea rather than the shorthand macula to focus on a unified set of mechanisms that include xanthophyll transport. The 3 mm-diameter region is preferentially used for clinical diagnosis of AMD using OCT.¹⁹ A 6 mm-diameter region sometimes called the macula should be called the central area. 

Polyak and the ETDRS Grid

Neurologist Stjepan Poljak, MD (1889–1955; a.k.a. Stephan Polyak), published notable and highly cited English-language textbooks on retinal neuroanatomy.^20,21 He expanded cellular neuroscience of Cajal with original photomicrographs of individual retinal neurons and glia stained by the Golgi method. He defined cell types, layers, and retinal regions with distinct cellular content.^13,22 Foveal specimens of humans and nonhuman primates are complete and of high quality. Aesthetic figures, detailed tissue preparation methods, and extensive literature reviews are provided. These definitive texts nevertheless lack information about RPE and choroid, as well as age, sex, race, and provenance of the tissue samples. 

Polyak defined concentric retinal regions aligned on the central bouquet of thinnest foveal cones. The central area (area centralis in other vertebrate species) has a continuous layer of ganglion cells, except at the foveal pit, and includes a layer of angled Henle fibers (photoreceptor axons and Müller glia). The outer diameter of the central area can be now determined accurately for individual eyes using imaging technology that reveals thickness of the ganglion cell layer, presence of the Henle fiber layer, and visibility of faint yellow. Herein, we use a 6 mm-diameter central area (rather than Polyak's 5.5 mm diameter), because the AMD risk profile for the 6 mm diameter region is known from epidemiology. 

We revisit Polyak's zones to define regions relevant to AMD onset and progression (Tables 1, 2) in the plane of photoreceptor inner segments. Within the central area, the internal fovea (the pit) has a diameter of ∼0.8–1.5 mm and contains an all-cone foveola (350 µm in diameter) (Table 1). The parafovea is a 1.0 mm-wide annulus (0.5–1.5 mm, inner-outer radius) surrounding the fovea, where rods intrude between cones. The perifovea is a 1.5 mm-wide annulus (1.5–3.0 mm inner-outer radius) in which cones are surrounded by two rings of rods. The near periphery is the next annulus, with rings of rods (two- to three-wide) between cones. Rods are more densely packed in regions further from the fovea.¹⁸ 

Table 1.

 

Cell Densities and Ratios in Subfields of a Supplemented (s)EDTRS Grid

 

Table 2.

 

Photoreceptor and RPE Numbers and Densities in Fovea-Centered Regions

 

The 6 mm-diameter ETDRS grid can link AMD-relevant findings in histology, molecular assay, clinical presentation, and epidemiology. The original grid was devised seemingly without neuroanatomic considerations because the width of veins at the optic nerve head could serve as internal rulers in fundus photographs.²³ Nevertheless, grid subfields fortuitously match the Polyak regions (Tables 1, 2) making the ETDRS grid a convenient tool for assessing the neural retina. We suggest adding a new ring (near periphery [NP]) to include areas of high rod density in a supplemented ETDRS (sEDTRS) grid. Of two OCT devices widely used for AMD management and research, the Spectralis (Heidelberg Engineering, Heidelberg, Germany) review software uses a 6 mm-diameter grid, and the Cirrus (Zeiss, Oberkochen, Germany) uses a 5-mm-diameter grid. Both include a central 3 mm-diameter zone. Thus the possibility of standardization on the macula lutea is good. 

Human Photoreceptor and RPE Topography

Visualizing and Quantifying Photoreceptors

High-quality light microscopy of photoreceptor inner segments in the human retina (Fig. 1)²⁴ propelled single-cell clinical imaging via adaptive optics-assisted scanning laser ophthalmoscopy. Flat mounts enable accurate and unambiguous counts of inner segments, which constitute the retinal image plane without requiring corrections for cells split by sectioning.²⁵ Because of computer-assisted microscopy and reconstruction methods, this work also provided complete digital two-dimensional maps of rod and cone density, allowing photoreceptor abundance to be treated as a continuous variable. These maps replicated key aspects of the single eye of a 16-year-old male analyzed by Østerberg in 1935,²⁶ such as the high density of foveal cones and the rod-free zone. Complete maps showed new aspects, such as the overall rod dominance of human retina, a ring of high rod density surrounding the fovea and the optic disk, with a hot spot in the superior retina, near the arcades, plus a streak of high cone density aligned along the horizontal meridian. The later discovery of clinical features that appeared aligned with the rod ring (i.e., SDD and high signal in fundus autofluorescence imaging) motivated the proposed sETDRS grid. 

Figure 1.

 

Human photoreceptor and RPE morphology varies with location. (A, B) Mosaic of photoreceptor inner segments at the ellipsoid level, captured by video-enhanced differential interference contrast microscopy. Fovea has only cones. Perifovea has 3-fold larger cones surround by complete rings of rods. (C, D) RPE cytoskeleton labeled by Alexa 647 tagged phalloidin, captured by epifluorescence microscopy. RPE cells have a precisely polygonal cytoskeleton with straight edges and sharp vertices. (E, F) Autofluorescence emissions (after 488 nm excitation) in the same cells as in panels C and D. The fovea has larger cells, a higher proportion of hexagonal cells and weaker autofluorescence emissions at this wavelength (C, E) than cells in the perifovea (D, F). Tissues are from white donors, >80 years of age. Scale bar: 10 µm, applies to all panels.

Visualizing and Quantifying RPE

Microscopy of human RPE supports fundus autofluorescence imaging, which in turn enables the visualization of RPE health and metabolism. Figure 1 shows RPE cells at the fovea and perifovea from a study²⁷ designed to emulate the photoreceptor maps. Unlike cones and rods, which share the retinal image plane, the RPE forms a continuous layer of one cell type. Numerous histological studies of RPE did not specify the exact position of the fovea in their specimens.^28–30 The largest study to date on human donor flat-mounts with precise information on fovea has shown how RPE appearance varies depending on retinal localization, age, and disease status.²⁷ The highest RPE cell density is found at the fovea, with decreasing numbers at the perifovea and near-periphery.²⁷ In contrast to the distribution of photoreceptor abundance over the same retinal distances, the RPE exhibits modest differences in density. Both in the perifovea and periphery, RPE cells enlarge. Although RPE cellular geometry at the fovea is mostly hexagonal with six neighbors, more cells have five or seven neighbors toward the near periphery. These regional differences in cell area were also recently confirmed and extended in a flat-mount study using AI-based image analyses to characterize five different RPE subpopulations. Regions were oriented concentric to the fovea extended to the ora serrata (retinal edge), with zone P1 corresponding to the macula lutea.³¹ 

An RPE hallmark is accumulation of autofluorescent inclusion bodies (often called granules) within the cell body, starting in fetal life and increasing through adolescence until at least age 70 years.^32,33 Although the individual fluorophores are not yet known, the majority (if not all) of the autofluorescence is widely attributed to bisretinoid byproducts of the visual cycle.³⁴ These moieties are stored in lipofuscin and melanolipofuscin, two types of membrane-bound organelles of lysosomal origin within RPE cell bodies. Super resolution microscopy (structured illumination) was used to demonstrate that in foveal RPE cells, melanolipofuscin is the abundant granule type, whereas lipofuscin is the dominant type in RPE cells at the perifovea and near-periphery.³⁵ This difference in granule distribution also explains clinical fundus autofluorescence with blue exciting light showing a weak signal at the fovea (in addition to light absorption by xanthophyll pigment) and highest signal in the sETDRS outer and NP rings. RPE cells at the latter locations have the highest number of lipofuscin granules.³⁶ 

The above-mentioned autofluorescent properties hold for excitation of lipofuscin and melanolipofuscin fluorophores with short-wavelength lights. At the other end of the spectrum, red and near-infrared light excite other RPE pigments, namely melanin in melanosomes and melanolipofuscin,³⁷ as currently investigated at the subcellular level.^35,38 Results will elucidate the significance of these fluorophores for RPE metabolism and clinical near-infrared fundus autofluorescence.³⁹ 

Photoreceptors and RPE in Zones of sETDRS

Figure 2 shows spatial density of outer retinal cells (cells/mm²) along a horizontal slice across the sETDRS grid through a young adult retina. Cones/mm² exhibit a sharp peak and an approximately fivefold declining gradient with eccentricity across the central subfield. Rods/mm² go from 0 to 130,000 just outside the grid, in the near periphery. The rod-to-cone ratio increases steadily from 0 where the rods first appear (at 175 µm from the foveal center) to ∼25 at the edge of the sETDRS grid. The highest values of this ratio (∼30) are attained 8 to 10 mm from the foveal center because of continued fall-off of cones. In the central area, densities along meridians may differ, with <15% higher densities at the same eccentricities on the horizontal meridian than the vertical meridian. Despite these asymmetries, including a hotspot of high rod density superior or superior-temporal to the fovea, meridional effects are small relative to eccentricity effects (5 ÷ 0.15 = ∼33). 

Figure 2.

 

Cell densities and ratios in human central retina. (A) Spatial density of cones, rods, and RPE cells along the horizontal meridian of young adult human retina. Rods and cone inner segments were counted in the same flat-mounted tissues; ratios were calculated directly.²⁴ RPE was counted in separate eyes using similar methods.²⁷ Photoreceptor: RPE ratios were computed by matching eccentricities and by assuming for simplicity that each RPE cell tends to the photoreceptors directly above it.⁴¹ The grading grid of the ETDRS²³ is at the right. Outer diameters of the outer ring, inner ring, and central subfield are 6, 3, and 1 mm, respectively, and are not drawn at the same spatial scale as the main graph. Proposed sETDRS includes an additional ring with inner and outer diameters of 6 and 9 mm, respectively, encompassing the near-periphery region. Eccentricities on the graph and rings in the grid are color-coded to match the concentration of xanthophyll carotenoid pigment, shown in projection view in panel A and cross-sectional view in panel B. Central subfield and inner ring together comprise the macula lutea and exhibit the highest and next-highest population-level risk, respectively, for AMD progression.⁸⁹ ONH, optic nerve head. (B) Schematic cross-section of a human fovea, with vascular plexuses and xanthophyll pigment indicated. Xanthophyll carotenoid pigment (yellow) is schematized¹⁴ from microdensitometry of sections through a macaque monkey fovea.²²⁴ GCL, ganglion cell layer; HFL, Henle fiber layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS/OS, inner and outer segment of photoreceptors together; ONL, outer nuclear layer; OPL, outer plexiform layer; OS, outer segment; RPE, retinal pigment epithelium; SCP, superficial capillary plexus; DCP, deep capillary plexus; ICP, intermediate capillary plexus; ChC, choriocapillaris. A conversion factor of 0.288 mm/deg of visual angle is used.

Figure 2 also shows that the photoreceptor-to-RPE ratio, a measure of the degradative load on RPE, exhibits complex relationship with eccentricity. This ratio is high (∼29) at the foveal center where cones are numerous. Counterintuitively, it is lowest (10–12) within the perimeter of the rod-free zone (i.e., where rod density is zero and cone density is reduced from its peak to the lowest point on the retina). From this nadir, the photoreceptor-to-RPE ratio steadily increases to >30 in the NP ring and then decreases at further eccentricities (not shown). A similar relationship was described in 2002 for non-human primates.⁴⁰ Calculation of the degradative load assumes that each RPE cell services only those photoreceptors immediately above it. Recent volume electron microscopic reconstruction has shown notable branching of apical processes, suggesting that some outer segments may be contacted by more than one RPE cell.⁴¹ 

Table 1 shows cell densities, ratios, and terminology in Polyak regions and proposed sETDRS equivalences of use for imaging and vision studies, in young and aged donor eyes with normal central retinas.^24,27,42 We report only sETDRS central and annular subfields (Table 1) and circular regions (Table 2) and not also individual sectors or quadrants. We retained the inner-outer ring names because of their history of usage. Meridional effects are negligible for inner and outer ring, as noted above, but the NP ring includes the optic nerve head region where photoreceptors are absent or sparse^24,43 (see notes to Table 1). In the sETDRS scheme, the fovea is the cone-rich central subfield, with a rod-to-cone ratio of 0.3 (young) and 0.4 (aged). The ratio is non-zero because of scattered rods that appear near the subfield perimeter. The parafovea is the sETDRS inner ring, with a rod-to-cone ratio of 4.2 (young) and 3.6 (aged) and thus is rod-dominated. The perifovea translates to sETDRS outer ring, with a rod-to-cone ratio of 12.9 (young) and 10.9 (aged), thus appearing very rod dominated. A proposed NP (near periphery) ring with inner-outer radii of 3 and 4.5 mm, respectively, is highly rod dominated with a ratio of 19.8 (young) and 16.3 (aged). The region of maximal rod-to-cone ratio (>30) is at 8–10 mm eccentricity (not shown). Degradative load (photoreceptors-to-RPE) in the four sETDRS regions is lowest in the central subfield (young-aged, 11.9–11.0, 17.3–13.2, 28.7–21.4, 32.8–26.2, for central, inner, outer, and NP subfields, respectively). Note that the peak of cone density and high degradative load in that one small spot (Fig. 2) is not captured by averaging across the central subfield in this manner. 

Table 2 shows the cumulative number of cells and mean density within fovea-centered circles of increasing diameter for use in vision tests of mean regional sensitivity and in molecular studies using circular tissue punches. Each ring is larger than the one internal to it. In a cumulative measure, any one circular region is dominated by the cell populations in the largest ring it contains. Thus the descriptive terms for each region are taken from those applied to the largest contained ring. One important message is that the macula lutea (central subfield + inner ring) is rod-dominated (rod-to-cone ratio, 3.1 in young adults, 2.6 in older adults). 

Table 3 compares young and aged directly via ratios of cell densities in sETDRS subfields and accumulated numbers within circular regions (from Tables 1 and 2, respectively). Smaller ratios of aged to young indicate greater loss in aging. These illustrate that slightly different conclusions can be drawn using these geometries. The anatomic region with the strongest aging effect is the circular macula lutea (bottom of Table 3), where aged eyes have 24% lower rod numbers than in young eyes (i.e., rod-to-cone ratio = 0.76). Numerically, more rods are lost in sETDRS NP ring in aging than in the sETDRS inner ring because the overall cell densities are higher. However, cell loss in NP ring is proportionately less than in the sETDRS inner ring. We focus on macula lutea (Polyak's fovea + parafovea, sETDRS central subfield + inner ring). 

Table 3.

 

Ratios of Outer Retinal Cell Populations in Young Versus Aged Normal Eyes

 

Huge eccentricity-dependent differences in photoreceptor populations represent a large dynamic range for an independent variable in a small space. This increases the value of precisely placed assays, whether in living people via imaging or in postmortem specimens. Conversely, inattention to the distance of assay locations from the fovea potentially adds noise to measurement data that is worse at low eccentricities close to the high-risk area. The AMD-vulnerable macula lutea is only 0.7% of the total area of a 1000 mm² human retina. Assays such as extractions of whole eye cups and biopsies of aqueous and vitreous need to be designed and interpreted accordingly, ideally within information about the extramacular areas if possible. Furthermore, in a 6 mm-diameter tissue punch typical for molecular studies of human retina,⁴⁴ a 3 mm-diameter macula lutea is only 25% of the area. 

Of two essential laboratory animal species for AMD research, Old World non-human primates have foveas, and they can develop soft drusen.^45,46 Monkeys have not yet been shown to progress to either atrophy or neovascularization. Interestingly macaque monkeys have overall fewer rods than humans,⁴⁷ not only because of their smaller eye size but also overall lower rods/mm². In contrast, mice have highly rod-dominated retinas lacking a fovea, with an overall rod-to-cone ratio higher than the NP ring in humans (Table 1). Mice have no equivalent to the vulnerable parafoveal area of low rod-to-cone ratio in humans. Publications concluding otherwise misrepresented the human photoreceptor distribution,^48,49 a misconception that this review may address with Tables 1 and 2. For focused explorations of key pathways, mice remain excellent models. 

A Multilayer View of Aging Relevant to OCT

Fovea in Cross-Section

The bottom panel of Figure 2 shows a cross-section of human fovea scaled to the sETDRS grid, with the superficial, intermediate, and deep capillary plexus of the retinal circulation and the choriocapillaris of the choroidal circulation. The yellow macular xanthophyll carotenoids are concentrated in the foveal center, and along the two plexiform layers and nerve fiber layer. This pigment comprises two polar xanthophyll carotenoids of dietary origin lutein and zeaxanthin, plus an intraocularly produced metabolite meso-zeaxanthin. All are concentrated in the 3 mm-diameter macula lutea, with zeaxanthin highest in the foveal central bouquet.^20,50 Xanthophylls intercalate into lipid bilayer membranes,⁵¹ potentially affecting functions of Müller glia, photoreceptor axons, and rod- and cone-driven neural circuits. 

Separating Three-Layer Bruch's Membrane From Basal Laminas

In addition to precision in the x,y plane, a new level of precision is available in the z-axis, thanks to OCT and recent histology that confirmed and extended many aspects of foundational work by Sarks et al.^52–55 (Supplementary Materials for a bibliography). Figure 3 shows 10 layers of distinct biology between the external limiting membrane of the neurosensory retina and the choriocapillaris, a specialized microvasculature of systemic circulation. 

Figure 3.

 

Deposit-driven AMD in histology and OCT. The vertical dimension is expanded to highlight ten anatomic layers between the external limiting membrane (ELM) and choriocapillaris (ChC) endothelium. Columns at the left and right show the outer retinal reflective bands of spectral domain optical coherence tomography.²²⁵ Vascular BrM⁶⁴ consists of the three middle layers: inner collagenous (ICL), elastic (EL), and outer collagenous (OCL). Both RPE and ChC endothelium rest on basal laminas (BL). BLamD is a stereotypically thickened extracellular matrix (green, in middle and right) between the RPE plasma membrane and RPE-BL in AMD (see also Fig. 4). Soft druse material is continuous with BLinD in the same sub-RPE-BL compartment. It is also found in basal mounds within BLamD.^55,64 Between the RPE and photoreceptors is subretinal drusenoid deposit (SDD; first called reticular pseudodrusen), stereotypic extracellular material that is reflective on OCT and directly disrupts photoreceptors.²²⁶ High-risk soft drusen material is a direct precursor to type 1 macular neovascularization (MNV, up arrow) in the sub-RPE-BL space. SDD is a risk indicator for type 3 MNV (down arrow) (first called retinal angiomatous proliferation),^227,228 of retinal origin and typically developing closer to the fovea than SDD themselves. Figure is slightly modified from source from Chen et al.¹¹ OS, outer segments of photoreceptors; M, melanosome; ML, melanolipofuscin; Mt, mitochondria; RPE-BL, RPE basal lamina; ChC-BL, ChC basal lamina.

Both Gass⁵⁶ and Sarks et al.⁵⁷ opined that AMD pathology can be best understood if BrM is considered to consist of only three layers (inner collagenous, elastic, outer collagenous) (Fig. 4) rather than the five-layer structure of textbooks (including chapters by Curcio and Johnson^58,59). These three layers can be considered vascular BrM, having molecular, structural, and functional commonality with an arterial intima and the transport function of a vessel wall. A three-layer BrM incorporates high-resolution histology showing that type 1 macular neovascularization (of choroidal origin) ramifies in the sub-RPE-BL space.^11,60,61 This concept differs from within-BrM neovascularization inferred from lower resolution tissue and microscopy studies.⁶² 

Figure 4.

 

Redefining Bruch's membrane, a unique vessel wall. Bruch's membrane underlaying central retina from human eye donors of indicated ages. RPE is at the top and choriocapillaris is at the bottom. RPE basal lamina (BL, arrowheads, thick light blue), elastic layer (EL, yellow arrows, discontinuous in central area), and the choriocapillaris endothelium BL (thin light blue) are shown. Age 17 years: RPE has basal infoldings. Electron-dense amorphous debris and lipoproteins are absent from three-layer Bruch's. Bar: 1 µm. Age 46 years: Electron-dense amorphous debris and presumed lipoprotein particles are present. A coated membrane bounded bound (green arrowhead) contains lipoproteins. Basal infoldings are not visible. L, lipofuscin. Age 65 years: Basal laminar deposit (green, with asterisk) has formed between the RPE basal surface (not shown) and the native RPE basal lamina, either incorporating or replacing the basal infoldings. Membranous debris, also called lipoprotein-derived debris (red arrow) has electron-dense exteriors within BLamD. The sub-RPE-BL space is a potential space between the RPE -BL and the inner collagenous layer which may accumulate drusen, type 1 macular neovascularization (T1 MNV), and exudative fluid. Electron-dense amorphous debris and lipoproteins are abundant within the three-layer Bruch's membrane. Choriocapillaris BL was not visible at this location. Inspired by Curcio and Johnson.⁵⁹

A three-layer BrM also effectively assigns the two basal laminas (RPE and choriocapillaris) to the cells that synthesize them. It accommodates the pathologic appearance of basal laminar deposit (BLamD), a defining feature of AMD presence and severity.^63,64 Accordingly, mutations in monogenic disorders may lead to either BLamD (TIMP3, EFEMP1, C1QTNF5) or calcification of the elastic and collagenous layers (ABCC6). These genetic associations contrast with those for the complement system, for which gene products localize exquisitely to the sub-RPE-BL space in addition to BrM.⁶⁵ Complement factor H–related proteins localize precisely to the intercapillary pillars.⁶⁶ 

Big Effects in Aging Outer Retina

The vertically organized and tightly integrated physiological unit of photoreceptors, Müller glia, RPE, and choriocapillaris form the outer retinal neurovascular unit.⁶⁷ This unit, as conceived originally for the brain and then inner retina,^68,69 comprises microvessels, neurons, glia, pericytes, and extracellular matrix that couple blood flow to metabolic demands of neurons. AMD may be either a primary neurodegeneration or a disease of vascular/ metabolic insufficiency, depending on the initial site of damage. In AMD, an extracellular deposit (drusen) is the largest population-level risk factor. Of recognized molecular hallmarks of aging,^70,71 breakdown of intercellular communication and materials transfer thus seem especially relevant, as well as biological purpose(s) of communications (e.g., constitutive vs. response to injury). 

To prevent AMD onset and early progression, it is fruitful to identify and focus on major effects of healthy aging on outer retina. To link cellular mechanisms to OCT-anchored multimodal imaging, we compared neural (photoreceptors and RPE) and vascular layers (BrM, choriocapillaris, choroid) in seven published studies of aging in normal human central retinas. We fit lines through published data, with a simplifying assumption of monotonic change between 20 and 90 years. We show a line connecting the fit data for these two ages in Figure 5. References and details for these calculations are in the Supplementary Materials. This analysis assumes that layers with greater changes in aging are more likely to convert to frank pathology than layers with minimal change. 

Figure 5.

 

Aging in tissue layers vulnerable to AMD: large and small effects. Tissue-level measures of aging in normal human central retinas are compared by fitting lines through published data and connecting the fit data for age 20 years and 90 years, assuming a linear change over that period. The slope of the line for each feature is indicated. All measured were determined in laboratory studies except for choroidal thickness, which was measured in vivo by OCT. For references to data sources and assumptions behind the calculations see Supplementary Materials. (A) Rods decline 36%, cones decline 8%, and the RPE exhibits stable cell numbers between 20 and 90 years. (B) Histochemically detected esterified cholesterol (EC) in BrM rises 15-fold in normal aging and is the largest aging change reported to date. Thickness of three-layer BrM doubles in the same eyes. (C) Choriocapillaris density (ChC) declines 35% and choroidal thickness, which includes the choriocapillaris, 24%.

Figure 5A shows a 36% decline in the spatial density of parafoveal rods and a minimal 8% decline in foveal cones, measured in the same eyes. RPE number in the central area is stable throughout adulthood across three histologic studies with large sample size. This remarkable stability is supported by a recent massive population-based OCT study showing consistent thicknesses of RPE cell bodies across the ETDRS grid of >7,000 participants aged 30–95 years.⁷² Figure 5B shows a twofold thickening of three-layer BrM and a 15-fold increase in histochemically detectable esterified cholesterol, a core lipid of RPE-originated lipoprotein particles that constitute a large component of high-risk drusen. Figure 5C shows two aspects of vascular physiology, thickness of a macrovasculature (choroid) and coverage of BrM by a microvasculature, the choriocapillaris endothelium. These both change to a similar magnitude over adulthood as the rods, with choriocapillaris slightly worse than choroid. 

Because RPE has demanding dual roles maintaining photoreceptors and choriocapillaris, its status in healthy aging is paramount. One interpretation of Figure 5 is that rod photoreceptors do less well than RPE and with similar change as choriocapillaris coverage because rod inner segment mitochondria are further away from the choriocapillaris (∼50 µm) than mitochondria in the basal RPE (2–4 µm). The RPE can sustain photoreceptors even over neovascularization if proximity and vascular functionality are maintained and exudation is avoided.^61,73 This inference, if confirmed in further research, suggests that AMD is a vascular-originating disease. The combination of cellular-level clinical imaging and advances in molecular techniques such as single-cell RNA sequencing⁷⁴ make it possible to test this hypothesis in a robust manner. 

A Small Area at Highest AMD Risk Aligns With the Macula Lutea

Using Macular Topography To Dissect Mechanisms

Despite the limitations of color fundus photography in revealing AMD pathology, this technology was used for foundational population-based epidemiology studies that remain invaluable resources. These documented now well-accepted AMD risk factors such as aging and smoking and found intriguing results on plasma HDL that are still explored today in a genomic context.^75,76 Fundus features including drusen of defined dimensions and appearance plus pigmentary abnormalities were quantified within the overall ETDRS grid at five-year intervals, using the Wisconsin Age-related Maculopathy Grading System and similar systems.^77–79 These longitudinal studies showed that phenotypic changes were far more potent predictors of progression (i.e., higher odds ratios) than any combination of lifestyle risk factors and genetic variants.^80,81 By analyzing features within ETDRS subfields, it is possible to further link findings to cells of the overlying neurosensory retina. Surprisingly, such a regional analysis was done only several times, likely because of its time-consuming nature before the advent of digital image processing software, as well as the absence of clear-cut motivating hypotheses. 

Beaver Dam Eye Study Baseline

In 1996 the Beaver Dam Eye Study lead by R. E. Klein reported for >4,500 right eyes baseline frequencies of AMD presence in ETDRS subfields, weighted for subfield area.⁸² Thus, frequencies of 5.3%, 6.9%, and 5.1% for central, inner ring, and outer ring became 5.3%, 0.9%, and 0.2% after weighting. A ratio of the central subfield and the inner ring to the outer ring was 26.5 and 4.5, respectively. Thus AMD pathology was found a remarkable 26.5 times more often in the central subfield than in the outer ring, and this frequency dropped sixfold within the extent of the inner ring. The authors did not explain why they analyzed by regions, and they did not pursue this line of inquiry. 

Spatial Dissociation Of Degeneration, Dysfunction, and Macular Biology

In 2002, the tight foveal distribution of AMD pathology in the Beaver Dam Eye Study (Fig. 6D) was compared to photoreceptor loss and vision dysfunction in aging and AMD, as well as biological aspects of healthy macula.⁸³ By then, it had been shown that histologic rod loss was worse near the fovea whereas cones were relatively preserved (Fig. 6B).^42,84 It had also been shown that steady-state rod sensitivity in aging and early AMD (staged by fundus grading) was worse near the fovea whereas cone sensitivity was relatively well preserved (Fig. 6A).⁸⁵ Regional differences in steady-state rod-involving sensitivity have since been replicated many times. As shown in Figure 6C, prototype fundus autofluorescence imaging of normal retina by Delori et al.^86,87 revealed that the distribution of macular xanthophyll pigments in healthy retinas, thought to be protective,⁸⁸ exactly tracked the topography of AMD pathology (high in the foveal center, decreasing with eccentricity). Conversely, the distribution of lipofuscin-attributable blue-light autofluorescence, considered harmful, increased with distance (Fig. 6C). This paradoxical spatial dissociation of retinal biology with AMD risk is discussed below. 

Figure 6.

 

Spatial relationships among retinal biology, vision, and AMD pathology. (A) Visual sensitivity loss for rods in aging and early AMD is worse near the fovea than further away, whereas cone sensitivity is relatively stable over the same range. (B) Rod number decreases in aging and AMD near the fovea whereas cone number is relatively stable over the same range. (C) Retinal indicators in healthy central retina include lipofuscin-attributable autofluorescence, previously thought to indicate AMD progression risk, is lowest at the fovea, and macular xanthophyll pigments, thought to confer protection from AMD, is highest at the fovea. (D) AMD pathology at baseline in the Beaver Dam Eye Study, weighted for ETDRS grid subfield area, is highly focused in the central subfield. Figure is modified from Jackson et al.,²²⁹ which provides methods for the calculations. Data sources for individual curves: visual dysfunction,⁸⁵ photoreceptor degeneration,^42,84 lipofuscin-attributable autofluorescence,⁸⁶ macular xanthophyll pigment,²³⁰ and AMD pathology.⁸²

Blue Mountain Eye Study Ten-Year Follow-Up

The Blue Mountain Eye Study from Australia, uniquely among large population-based studies, reported AMD prevalence stratified by ETDRS zones at 10-year follow-up in 2474 eyes⁸⁹; these longitudinal data supported the observations from the Beaver Dam baseline. The odds ratio for progression to late AMD at 10 years because of drusen at baseline within the ETDRS central subfield was again 26.5 (confidence intervals = 13.3–52.7). The risk conferred by aging is thus larger than all other population-level risk factors. For example, all risk variants of major AMD genes plus moderate to severe lifestyle factors together provide odds ratios of 26–35 (15.9–40.8 moderate, 21.8–56.4 severe).⁹⁰ Smoking confers progression risk with an odds ratio of 3.1 (2.1–4.6).⁹¹ Low risk in the outer ring (odds ratio for progression 1.2, 0.3–4.3) separated central soft drusen from the many other drusen in peripheral retina.^92,93 Thus the margins of a high-risk area are captured within the ETDRS grid. Finally, risk conferred by pathology in the inner ring (odds ratio 8.6; 4.1–18.2) was one third of that in the central subfield. Thus progression risk is extraordinarily concentrated within the central 3 mm-diameter macula lutea. 

AREDS/AREDS2

Support for a focal area of AMD vulnerability is found in the unparalleled resources of standardized color fundus photography established by the Age-Related Eye Disease Study (AREDS) and AREDS2 study groups. Because of the enrollment of older participants, including those with AMD, and pooling data from ETDRS central subfield and inner ring,⁹⁴ these data are not directly comparable to studies mentioned elsewhere in this section. However, within-eye topographic analyses are valid. From 1992 to 1998, AREDS enrolled participants (aged 55–80 years) in a randomized clinical trial to evaluate the effect of oral antioxidant supplements on progression to advanced AMD.⁹⁵ At 10-year follow-up (n = 3549 persons),⁹⁶ progression rates for eyes with large drusen (and lacking RPE changes) limited to the macula lutea or the sETDRS outer ring was 21.7% and 3.6%, respectively. From 2006 to 2008, AREDS2 enrolled participants (aged 50–85) to evaluate the effect of oral supplements including lutein and zeaxanthin on AMD progression.⁹⁷ After 10 years’ follow-up, drusen outside the ETDRS grid in 3624 eyes were compared to 544 eyes lacking these drusen and found to confer no independent risk after adjustment for known risk factors.⁹⁸ 

Vienna Drusen Study (OCT)

A recent study asked if the topography of soft drusen from studies based on color fundus photography was replicable by OCT, which allows better distinction among druse types.⁹⁹ Large soft drusen were defined as dome-shaped sub-RPE deposits with a horizontal diameter > 125 µm and indistinct borders. In 62 eyes of 44 patients with a follow-up time of up to 78 months, drusen volume in the central subfield was 24.6 times higher than in the outer ring, with a steep fall-off to the inner ring, remarkably like the results of population-based studies.^86,89 In addition, all drusen topographies matched well with the distribution of macular xanthophyll pigment in 60 normal-aged eyes of 30 persons determined by two-wavelength autofluorescence in another European-descent patient population¹⁴ and less well with cone density from histology.²⁴ 

Topographic Considerations Absolve Lipofuscin

Figure 6 shows that fundus autofluorescence caused by lipofuscin, long thought to play a decisive role in AMD pathogenesis by leading to RPE death,^100,101 was low in the foveal center and high at the edge of the EDTRS. In other words, its topography was opposite to that of AMD lesions revealed by population-based epidemiology. Subsequent histologic validation of clinical autofluorescence imaging has clarified this relationship.^102–104 It was striking that RPE cell density was found to remain stable with age from the fovea to the near periphery (Tables 1, 2, Fig. 5), despite overall higher lipofuscin and melanolipofuscin in older eyes than younger eyes, as determined by super-resolution microscopy.^35,105,106 One study that measured both cell number and histologic autofluorescence showed that despite significantly higher autofluorescence signal in aged eyes, no differences in RPE cell density or morphology from younger eyes were detectable.²⁷ 

As Figure 2 shows, the photoreceptor-to-RPE ratio, a measure of degradative load on RPE and the input of outer segments contributing to lipofuscin and melanolipofuscin, has a biphasic relationship with distance from the foveal center. It is lowest ∼0.3 mm from peak cone density. In comparison, over a narrow eccentricity range (0–1.5 mm, 0°–5.2°) that includes this nadir, the frequency of early AMD lesions peaks at the center and decreases markedly and monotonically (Fig. 6D). We should add that on a larger scale, the known bisretinoids have been shown by three analytic chemistry technologies to be five- to nine-fold higher in the equatorial retina (i.e., outside the sETDRS grid) than in the central retina.^107–109 Overall, topographic considerations at the aging-AMD transition do not support a causative relationship of autofluorescence signal sources in central retina with AMD progression. 

We emphasize that there are clear changes in the RPE of AMD eyes.¹⁰² Several independent clinical studies using quantitative fundus autofluorescence have shown that autofluorescence intensities decrease across the entire posterior pole, even at early and intermediate stages of AMD.^110–113 Histologically, this decrease in autofluorescence may relate to extrusion of autofluorescent granules from the RPE cells, as well as granule aggregation that leaves signal-free areas of cytoplasm.¹⁰² 

How Feeding the Fovea Throughout Life Leads to High-Risk Drusen

A drusen biogenesis model with strong multidisciplinary support aligns the 3 mm-diameter high-risk area for AMD with the macula lutea, the neural tissue dependent on dietary delivery of xanthophyll pigments (Fig. 7). In brief, high-risk drusen may form as the RPE takes up plasma lipoproteins as part of normal physiology and releases unneeded lipids in its own lipoprotein particles. These are prevented from transiting to circulation due to age-changes in choriocapillary endothelium and BrM. Accumulation of histochemically detectable lipids in sub-macular BrM is universal in normal eyes and begins in late adolescence.^114,115 These deposits directly presage type 1 macular neovascularization in the same compartment (Fig. 4).^11,60 Recently, histology and clinical imaging converged to show that high-risk drusen are direct precursors to atrophy as well,^{104,116–119} thus completing a path from evolutionarily shaped gene expression to clinical presentation. High-risk drusen have thus emerged from mystery and possibly even epiphenomena to recognized precursors of advanced disease. 

Figure 7.

 

Model of drusen biogenesis. Five steps are depicted. (1) Plasma HDL and LDL carrying lutein and zeaxanthin²³¹ are taken up at RPE receptors (scavenger receptor B-1 and LDL receptor).²³² (2) RPE extracts xanthophylls for transfer to retina via interphotoreceptor retinal binding protein²³³ and possibly others.²³⁴ Cellular xanthophyll binding proteins GSTP1 and StARD3 have been localized in cones.^235,236 Xanthophyll has been localized in retinal layers beyond those accounted for by cones and in places consistent with Müller glia, suggesting additional proteins with binding and transfer capability such as FABP5.²³⁷ (3) RPE constitutively releases unneeded lipids back to circulation in large lipoproteins containing apolipoproteins B and E. (4) Lipoprotein particles accumulate in BrM starting in late adolescence and build up through adulthood²³⁸ because of impaired transport across aging BrM (which becomes cross-linked) and choriocapillaris endothelium (which degenerates). These accumulate as soft drusen material, separating RPE from choriocapillaris and containing pro-inflammatory, pro-angiogenic peroxidized lipids in an atherosclerosis-like progression.^239,240 (5) Soft drusen material is a direct precursor of type 1 (choroid-originating) neovascularization.^11,60

The RPE expresses hallmark genes of a native secretor of large lipoprotein particles containing apolipoprotein B, including microsomal triglyceride transfer protein.¹²⁰ Lipoprotein particles are neutral lipid droplets solubilized for transport through aqueous media like plasma and interstitial fluid with a surface of phospholipid, unesterified cholesterol, and proteins specified by the genome. RPE thus resembles liver and intestine, the sources of atherogenic plasma lipoproteins,¹²¹ as well as heart, which releases lipoproteins to forestall lipotoxicity.¹²² These genes are also expressed in retinas of mice,¹²³ which are separated from humans by 90 million years of evolution.¹²⁴ In mice, RPE-specific microsomal triglyceride transfer protein deficiency leads to outer retinal degeneration,¹²⁵ indicating that assembly of apoB containing lipoproteins is essential for mammalian retinal health. 

Sarks et al.¹²⁶ stated that soft drusen were found only in the central area. This was demonstrated directly in an analysis of 415 micro-dissected drusen from 9 eyes of 7 donors with large macular drusen.¹²⁷ Soft drusen, basal linear deposit (BLinD), and basal mounds are three different forms of the same material called “membranous debris” by the Sarks. This material was called “lipoprotein-derived debris” by Curcio and colleagues due to the likely mechanism of formation, without excluding other mechanisms. Pre-BLinD is rows of seemingly intact lipoproteins in the sub-RPE-BL space (first called Lipid Wall¹²⁸) and precursor to BLinD and soft drusen. Basal mounds are soft druse material trapped within BLamD of generally older AMD patients with poorer acuity^55,64; more details about its life cycle may emerge with new OCT technologies. Soft drusen are distinguishable from histologic nodular and cuticular drusen, which are globular and pan-retinal, by size, ultrastructure, and uptake of in vivo dyes for angiography.¹²⁹ Pan-retinal drusen may be associated with complement gene risk variants and confer risk in abundance and over a long time.¹³⁰ 

One key point is that soft drusen material is thickest under the fovea, as determined in a high-resolution histology survey of 23 early-intermediate short post-mortem AMD eyes and 29 age-matched control eyes.^11,64 Seminal molecular surveys of drusen either sampled outside the high-risk area¹³¹ or combined drusen across the retina¹³² so that high-risk drusen represented a very small proportion of the total.¹¹ 

Another key point is that strong evidence supports transfer of diet-originated lipids being a main driver in drusen biogenesis. First, drusen can form and remodel under a macular hole, indicating that a proposed lipid cycling system (Fig. 7) does not depend on the neurosensory retina.¹³³ Second, outer segments are not required for the deposition of sub-RPE material containing lipids, apolipoproteins, and calcium by well-differentiated RPE in culture.^134–136 Third, the fatty acid composition of all lipid classes in lipoproteins isolated from human Bruch's membrane indicates enrichment of linoleate (18:2), abundant in plasma, rather than docosahexaenoate (22:6), abundant in outer segments¹³⁷; the same is true for esterified cholesterol extracted from human Bruch's membrane.¹³⁸ Fourth, recent data point to overall prodigious dietary needs of outer retina for foveal cone energetics¹³⁹ and for maintaining lipid bilayer membranes throughout the retina, as follows. Unesterified cholesterol is one membrane component that is quantifiable with dual isotope tracer technology. In species with different degrees of similarity to human cholesterol homeostasis, 22% (mice, low similarity) and 47% (hamster, high similarity) of unesterified cholesterol in retina originates from diet. In brains of these animals, ≤6% of unesterified cholesterol comes from exogenous sources.^140,141 Future studies of these and related animal models will help determine, at the level of RPE lipoprotein assembly, the quantitative contributions to drusen formation of lipids from diet, ingested outer segments, and endogenous synthesis. 

How RMDA Probes the High-Risk Area

Dark adaptation is the recovery of light sensitivity of the visual system to a decrease in ambient illumination. It can be studied by exposing a subject's retina to a bright flash of light (called bleaching), followed by a rapid transition to darkness. During this time, the retina is probed by near-threshold flashes of light, and the subject indicates when they are visible. This dynamic measure of retinoid resupply differs from steady-state sensitivity shown in Figure 6. Readers are directed to excellent reviews of dark adaptation and its underpinnings in the RPE-based retinoid cycle, as summarized by Nigalye et al.¹⁴² These include multidisciplinary historical perspectives from visual psychophysics, biochemistry, biophysics, and cell biology.^143,144 

A monumental treatise by Lamb and Pugh¹⁴⁵ detailed known molecular mechanisms of the retinoid cycle and formulated a quantitative relationship with vision in humans and laboratory animals. They proposed that one rate-limited kinetic process underlies the regeneration of rhodopsin, psychophysical dark adaptation, and recovery of rod photocurrent in electroretinography. The Lamb and Pugh central hypotheses (summarized in Fig. 8) were that lower sensitivity after bleaching results from the presence of opsin protein lacking the chromophore 11-cis retinal, and second, the rate of sensitivity restoration is directly proportional to the availability of 11-cis retinal that can bind to naked opsin. As summarized,¹⁴³ until 11-cis-retinal binds and rhodopsin is re-formed, this protein produces a desensitizing ‘‘equivalent light.” Cone photoreceptors respond to much higher light levels and much more quickly than rods. As summarized,^143,146 cones also receive retinoids via Müller glia and thus are not as vulnerable to aging- and AMD-related degradation of a retinoid resupply route from the circulation across the RPE-BrM-choriocapillaris complex. The cone component of dark adaptation is faster than that of rods and is thus more difficult to measure. 

Figure 8.

 

Retinoid re-supply route from circulation. Regeneration of visual pigment and light sensitivity is rate-limited by delivery of 11-cis retinal (cis RAL) from the RPE to opsin (Ops) in outer segments (OS). The 11-cis retinal and opsin form a molecule capable of absorbing photons. After meta-rhodopsin intermediates (M2 and others), all-trans RAL that remains non-covalently bound to opsin is reduced by all-trans retinol dehydrogenase for return to the RPE. Opsin is thus liberated to accept new 11-cis-retinal. Retinoids from diet are delivered from circulation to OS. At seven numbered steps, changes caused by aging or pathology may impair retinoid transfer: (1) extravasation and uptake of circulating vitamin A complexes through membrane specializations of the choriocapillary endothelium; (2) diffusion across or binding to Bruch's membrane; (3) receptor-mediated uptake of retinoids by RPE basal infoldings; (4) isomerization or oxidation of retinol, or both; (5) intracellular transport through RPE cell bodies and apical processes; (6) trafficking through the interphotoreceptor matrix (IPM); and (7) uptake or loading of 11-cis retinal onto the opsin molecule. Schematic is inspired by Lamb and Pugh.¹⁴⁵ IS, inner segment.

In 1993, Steinmetz and colleagues¹⁴⁷ showed that RMDA measured at 3° (∼0.86 mm) from the point of fixation markedly slowed in patients with early and intermediate AMD. This investigation was prompted by their finding histochemically detectable lipids in normal adult BrM.¹¹⁴ To explain the visual deficit, these authors postulated a localized shortage of retinoids at the photoreceptors. This in turn was attributed to impaired transport from circulation, either because of this hydrophobic barrier, metabolic insufficiency of the RPE, or both. Shortly after, RMDA was shown to slow steadily throughout adulthood in a study of 99 persons across seven decades, screened for retinal pathology with fundus photography grading.¹⁴⁸ 

Figure 8 shows that retinoid resupply could be impacted by aging or pathology at seven steps or more between the choriocapillaris endothelium and outer segments, thus expanding the original hypothesis: (1) extravasation and uptake of circulating vitamin A complexes through membrane specializations of the choriocapillaris endothelium^149–151; (2) diffusion across or binding to BrM¹¹⁵; (3) receptor-mediated uptake of retinoids by RPE basal infoldings¹⁵²; (4) isomerization or oxidation of retinol or both¹⁵³; (5) intracellular transport through RPE cell bodies and apical processes¹⁵⁴; (6) trafficking through the interphotoreceptor matrix¹⁵⁵; and (7) uptake or loading of 11-cis retinal onto the opsin molecule.¹⁵⁶ 

Many of these steps are now or soon will be visible in living people through multimodal imaging, such as structural OCT (cross-sectional and en face) with 7 or 3 µm axial resolution, two-photon autofluorescence, indocyanine green angiography, and OCT angiography.^157–163 Thus the prospects of determining the quantitative contributions of individual steps to slowed RMDA in large samples of clinically characterized patients are good. Recent OCT data implicate structural degradation of RPE uptake and transfer interfaces at the basal infoldings (step 3)¹⁶⁰ and apical processes (steps 5–7).^157,158,164 

Many stimulus factors impact RMDA measurement,¹⁴² so within-eye tests of locations with differing neural compositions under the same bleaching conditions represent well-controlled, informative experiments. Abundant and collectively robust evidence indicates that RMDA is impaired close to the fovea, where the rod-to-cone ratio is low (<4, Fig. 2). Starting with the article by Haimovici et al.,¹⁶⁵ studies of 20 to >500 aged persons with and without early or intermediate AMD^166–172 have shown that RMDA is more markedly delayed close to the fovea (3°–5°) than it is at 10° to 12° from the fovea (Table 4). A key finding for the field was that eyes with SDD, which start where rods are abundant,^7,173 have particularly poor RMDA.¹⁷⁴ Even in eyes with SDD, however, RMDA is slowest near the fovea and not at the deposit location (Table 4). Delayed RMDA is associated with structural changes in the outer retinal bands of OCT at 0.5 mm (1.7°) and not at 2 mm (6.9°)¹⁵⁷ in normal and AMD eyes not screened for SDD. An important question is when during the lifespan regional differences arise. A study of normal eyes of persons 28.9 ± 3.5 years versus 67.3 ± 10.7 years (n = 15 each) found regional differences in both groups¹⁶⁷; this intriguing result warrants investigating with a larger sample. 

Table 4.

 

Rod-Mediated Dark Adaptation Slows More Near the Fovea Than Far Away

 

Cone Resilience and Rod Vulnerability, a Center-Surround Model

We interpret this regional effect in rod-mediated vision as occurring on the edge of a high-risk area created by lifelong sustenance of foveal cones. As illustrated in Figure 9, spatially concentric, opposing mechanisms can be assembled mathematically to form a narrow central region of positive-going influence atop a broader and shallower surround of a negative-going influence. The figure plots xanthophyll carotenoids as a focused positive-going center of help to cones and their support system (Fig. 9A). Drusen and their precursors are shown a broad negative-going valley of harm affecting the support systems of cones in the center and rods around the edge (Fig. 9B). The valley of harm is broad because xanthophylls extend beyond the foveal center into the plexiform layers, where Müller glia interact with bipolar and ganglion cells. 

Figure 9.

 

Center-Surround model of cone resilience and rod vulnerability. Aging and AMD can be modeled as difference of 2-dimensional Gaussian surfaces.^176,177 In the top row is an en face view of help via macular xanthophyll pigment (orange) and harm via soft drusen/basal linear deposit and sequelae. In the bottom row help and harm are plotted on one vertical axis, in positive and negative directions, respectively. (A) The distribution of xanthophyll carotenoids, as shown in Figure 2, is a focused center of help in the macula lutea. (B) The distribution of soft druse material and sequela is shown as a broad circular area of harm. (C) Together, help and harm make a narrow center of foveal cone resilience on top of a broad surround of parafoveal and perifoveal rod vulnerability.

If we hypothesize that xanthophylls are transported from plasma to macula lutea as in Figure 7, then a wider region of RPE-BrM-choriocapillaris than the rod-free zone is required. There is some evidence that RPE under the macula lutea has a distinct lipid composition.¹⁷⁵ Combining the positive and negative effects creates a narrow center of foveal cone resilience amid a broader annular surround of parafoveal and perifoveal rod vulnerability (Fig. 9C). The result has a striking qualitative similarity to the annulus of photoreceptor degeneration and dysfunction in aging and early AMD. Rods are harmed as lipids accumulate in BrM, block RPE exchange with the choriocapillaris, and promote drusen formation, inflammation, and type 1 neovascularization. Cones themselves are largely spared until late AMD because of protection by Müller glia, a rich reservoir of xanthophylls and other essential support services. 

We further suggest that continuing progress in clinical imaging will eventually make this topographic relationship quantitative. Figure 9 employs a difference of Gaussian curves. This mathematical construct is familiar to vision science, because it was used to model excitatory centers and inhibitory surrounds of retinal ganglion cell receptive fields^176,177 (i.e., the part of the visual world to which individual neurons respond). Local concentration of xanthophyll carotenoids may be estimable with intensities from two-wavelength autofluorescence and retinal thicknesses from OCT.^16,178,179 Volumes of drusen at their earliest stages may be determinable automatically from large population-based studies of aging using OCT technology with adequate axial resolution and eye tracking for outer retinal pathology.⁷² 

Considerations for Visual Function Testing in Geographic Atrophy

The Foveal Center Is Variably Defined

An origin (0,0) of a polar coordinate system for retina has been variably defined, and this variability factors into how vision is assessed clinically. The smallest and most tightly packed cones connect to postreceptoral neural circuitry for high acuity photopic vision. We propose these cones as a biologically defensible origin. They are interleaved with Müller glia in a circular central bouquet of diameter 150 to 200 µm¹⁸⁰ or even smaller (100 µm)²¹ (see Supplementary content of the article by Litts et al.¹⁸¹). Variability in histologic estimates of peak foveal cell density owes in part to requirements for a comprehensive tissue visualization technology applied to very well-preserved specimens, viewed in an en face plane, for greatest accuracy. Our photoreceptor maps used specimens rigorously screened for foveal intactness and a 37 × 37 µm (0.13°) window to visualize cones densely packed as high as >300,000/mm².²⁴ For mapping RPE with a low peak density (up to 7400/mm²),²⁷ the center was determined by aligning photographs taken before and after removal of neurosensory retina, with confirmation by high pigmentation of foveal RPE, an inherently less precise method than possible for photoreceptors. 

By clinical OCT, long cones of the central bouquet can be recognized by an inward rise of the external limiting membrane and ellipsoid zone.^182,183 Others use minimum distance between the internal limiting membrane to RPE-BrM, within the foveal pit¹⁸⁴ or maximum distance between the ellipsoid zone to interdigitation zone, which captures part of the outer segment. The central bouquet in clinical OCT is associated with specific disorders¹⁸⁵ and may be viewed with adaptive optics assisted OCT.¹⁸⁶ Outer segment length in OCT has been assessed as a surrogate for peak cone density^187,188 with mixed results. A functional measure, the preferred retinal locus of fixation (PRL), centers visual field maps in microperimetry and can be registered to near-infrared reflectance (locator) images or preferably OCT volumes.¹⁸⁹ The PRL is also used to center the EDTRS grid in review software of OCT devices. 

Not all anatomic features (high cone density, inward rise of reflective bands, outer segment length, thinnest point in the foveal pit) may align vertically with the PRL in any one fovea.¹⁹⁰ Indeed, a provocative recent study suggests that the PRL is systematically offset from the locus of highest cone density in fellow eyes to maximize binocular high-acuity vision.¹⁹¹ Steep gradients of cone density within the central subfield may be captured by adaptive optics-assisted scanning laser ophthalmoscopy^187,192,193 in smaller populations to date than currently possible with OCT. The impact of errors in localizing in an origin because of biological variability may be small for outcome measures that are averaged over a large area (e.g., mean sensitivity in a 68-point microperimetry grid ¹⁹⁴). However, for mechanistically oriented imaging studies¹⁴ that invoke individual differences in foveal shape,^16,195 in turn resulting from genetically orchestrated developmental processes,^196–198 defining the center consistently and precisely can accelerate new insights. Today, the integrity of the central subfield is an important surrogate measure of photopic acuity, as developed in the next section. 

Are Cones Protected, Even in Atrophy?

The exact foveal center is of immense importance for the clinical consideration of geographic atrophy growth rates.^199,200 Recent phase 3 trials for complement inhibitors approved by United States Food and Drug Administration for treating geographic atrophy define lesion location and extent in the study protocols. For example, in the DERBY and OAKS pivotal trials for pegcetacoplan,²⁰¹ subfoveal and extrafoveal lesions were distinguished based on whether they occupied the foveal center point. They did not define lesions with respect to an area as large as the central subfield. Neither did the GATHER2 trial (for avacincaptad pegol) consider the central subfield.²⁰² A post-hoc OCT analysis of a small phase 2 trial for multi-action risuteganib showed much higher responder rates for eyes lacking atrophy in the central subfield than eyes with atrophy.²⁰³ Thus it is essential to critically examine and standardize how the fovea is defined in AMD trials and treatment decisions. 

The growth of geographic atrophy and how it impacts acuity depends on retinal location, underscoring the importance of neuroanatomy. Lesions that start outside the fovea may progress rapidly, then naturally slow down near the fovea. Recent meta-analyses by Shen et al.²⁰⁴ of eyes with atrophy in color fundus photography provided key insight despite the limited precision of defining the foveal center (± 200 µm) in this modality. One analysis (254 eyes) showed that atrophy growth towards the fovea is significantly less than growth away from it, particularly for lesions within 600 µm eccentricity²⁰⁴ (radial growth rates: fovea 40 µm/year, perifovea 100 µm/year). One factor potentially contributing to this sparing may be Müller glia, which survive the loss of photoreceptors²⁰⁵ and retain detectable xanthophylls.^206–208 Furthermore, another analysis (365 eyes) showed that photopic visual acuity declined dramatically in relation to the amount of the remaining 1 mm-diameter central subfield.²⁰⁹ Atrophy within this subfield²⁰⁹ negatively correlated with visual acuity, with 34.8 letters lost if the entire subfield was affected. Therefore knowledge of the exact location and extent of fovea tissue with the histological characteristics of photoreceptors and RPE described above is critical. 

The complement inhibitors recently approved by the Food and Drug Administration²⁰¹ could further decelerate growth, particularly within the fovea, and thus preserve or lengthen a period of good vision in late life. Currently, the use of pegcetacoplan is questioned by some because of the lack of visual benefit.²¹⁰ Of note, inclusion criteria for the pivotal phase 3 trials did not exclude patients with atrophy already in the central subfield.²⁰¹ We suggest that visual benefit might be more apparent if they had. Furthermore, even there was no benefit in visual acuity after 24 months’ follow-up ongoing extension studies²¹¹ could show that in the treated eyes in the long-term, visual acuity is maintained significantly longer than in nontreated eyes because of the protection of photoreceptors and RPE in the central subfield. These findings should be carefully acknowledged by other regulatory bodies and the treating physicians in patient education and selection. 

Strengths, Limitations, Future Directions, and Conclusions

Strengths and Limitations

This review considers retinal topography, as sculpted by evolution, as a decisive factor in AMD onset and early progression. We offer an sETDRS grid with a ring of highest rod densities along with newly re-computed photoreceptor and RPE cell densities from high-quality studies of retinal flat mounts. A pathogenesis model spatially unifies very large effects supported by multidisciplinary research extending over decades: the foveal singularity, precisely localized changes in aging, epidemiologic and molecular evidence for lipid-rich drusen under the macula lutea, and well-honed biophysics of dark adaptation. 

A limitation is that most studies of human eyes (in vivo and ex vivo) were confined to individuals of European descent; however, other populations clearly differ.^212,213 A limitation to the aging analysis (Fig. 5) is that cell number and tissue thicknesses may be perceived as crude measures; they are, however, quantifiable. A limitation to the regional analysis of color fundus photography-based epidemiology is that features seen by OCT are just being explored, and they confer large risk too.²¹⁴ Finally, this review is narrative and not systematic. Many ideas are speculative. However, they are testable, especially as imaging technology improves and as new AMD-relevant model systems allow experimental studies. 

Future Directions and Recommendations

We recommend using standardized anatomic language for human retinal regions and for describing the cell populations contained therein (Tables 1, 2). The 3 mm-diameter macula lutea is the region at highest risk for AMD progression, it is rod-dominated, and rods around the perimeter of this region are cogent functional indicators. We offer an integrated hypothesis based on soft drusen biology and transfer of dietary xanthophyll carotenoids and other key essentials to foveal cells to explain this regional propensity. Further clinical and histological studies, including high-resolution imaging, will lead to new definitions of structural and functional parameters in the macula lutea and the fovea within it, as well as rigorous testing of these hypotheses. New data on the dimension of the central bouquet of finest foveal cones could be used to supplement the sETDRS further. Consideration of the four-dimensional puzzle can help prioritize areas for investigation. We can anticipate convergence of tissue-level considerations with AMD genetic associations. A relationship of HDL genes with the xanthophyll bioavailability axis is promising,¹⁷⁹ as is exploration of complement genes in early AMD.^65,215,216 

Using the sETDRS central subfield (rather than a single point) highlights the region needed for photopic visual acuity in late disease. This subfield should be part of the analysis in clinical studies and therapeutic decisions in this new era of complement inhibitors. Using the sETDRS NP ring will facilitate mechanistic comparison with high-rod density regions to lower rod density regions in the vulnerable parafovea, of relevance to SDD's role in AMD progression. Mathematical modeling of AMD's beginning will be possible with population-based imaging of drusen and xanthophyll abundance to formalize and test the Center-Surround model. Retinal topography is an acknowledged challenge to engineered experimental model systems like organoids.²¹⁷ However, it seems possible to replicate just the vulnerable low rod-to-cone ratio area.²¹⁸ 

Conclusions

With population aging and a shrinking eye care workforce,²¹⁹ precision prevention of a disease affecting >200 million people worldwide should be prioritized. Cellular-level resolution in clinical imaging will achieve maximal utility if high-quality cellular, subcellular, and molecular data from human eyes can drive hypotheses to be tested with large cohorts. Using such data, we offer standardized anatomic language for consideration by the scientific community. The transition from aging to intermediate AMD is not yet incorporated in regulatory-approved outcome measures and treatment plans, but it could be because biological effect sizes pertaining to diet, drusen, and microvasculature are large, promising effective interventions. 

Acknowledgments

C.A.C. cherishes the impactful lives and far-reaching scientific contributions of two women of advanced vision and persistent dedication to human progress, Anita E. Hendrickson, PhD (d. 2017), and Shirley H. Sarks, MD (d. 2023). 

C.A.C. gratefully acknowledges the opportunity to discuss concepts developed for this review at the 2022 Ryan Initiative for Macular Research (“Why the macula” workgroup, James T. Handa, MD, and Demetrios G. Vavvas, MD, PhD, co-moderators). Workgroup members included: Catherine Bowes Rickman, PhD, Joseph Carroll, PhD, Dolly Chang, MD, PhD, Gregory Hageman, PhD, Aparna Lakkaraju, PhD, Phil Luthert, MBBS, Robert Mullins, PhD, Eric Pierce, MD, PhD, Austin Roorda, PhD, Jose-Alain Sahel, MD, Giovanni Staurenghi, MD, David S. Williams, PhD, Carol Berkower, PhD. 

The aging analysis of Figure 5 was initially prepared for the Robert M. Boynton Lecture at the 2021 Fall Vision Meeting of Optica. 

Supported by R01EY029595 (C.O., C.A.C.) and R01EY027948 (C.A.C., T.A.); P30EY03039 (C.O.); Dorsett Davis Discovery Fund and Alfreda J. Schueler Trust (C.O.); and unrestricted funds to the Department of Ophthalmology and Visual Sciences (UAB) from Research to Prevent Blindness, Inc., and EyeSight Foundation of Alabama. Also supported by Federal Ministry of Education and Research PACEtherapy 01EJ2206B (T.A.). Photoreceptor and RPE topography studies were supported by NIH grant R01EY06109 and for RPE, also German Research Foundation DFG (Bonn, Germany) no. AC265/1-1. 

Disclosure: C.A. Curcio, Genentech/Hoffman LaRoche (F), Regeneron (F), Heidelberg Engineering (F), Apellis (C), Astellas (C), Boehringer Ingelheim (C), Character Biosciences (C), Osanni (C), Mobius (C); D. Kar, Apellis Pharmaceuticals (E); C. Owsley, Johnson & Johnson Vision; K.R. Sloan, None; T. Ach, Roche (C), Novartis (C), Bayer (C), Nidek (F)