Fundus Autofluorescence Variation in Geographic Atrophy of Age-Related Macular Degeneration: A Clinicopathologic Correlation

Abstract

Purpose: The purpose of this study was to develop ground-truth histology about contributors to variable fundus autofluorescence (FAF) signal and thus inform patient selection for treating geographic atrophy (GA) in age-related macular degeneration (AMD).

Methods: One woman with bilateral multifocal GA, foveal sparing, and thick choroids underwent 535 to 580 nm excitation FAF in 6 clinic visits (11 to 6 years before death). The left eye was preserved 5 hours after death. Eye-tracked ex vivo imaging aligned sub-micrometer epoxy resin sections (n = 140, 60 µm apart) with clinic data. Light microscopic morphology corresponding to FAF features assessed included drusen-driven atrophy, persistent hyperautofluorescence (hyperFAF) islands and peninsulas within atrophy, and hyperFAF and hypoautofluorescence (hypoFAF) inner junctional zone (IJZ) and outer junctional zone (OJZ) relative to descent of external limiting membrane (ELM). Atrophy growth rate was calculated.

Results: HypoFAF atrophic spots appeared in association with drusen, and then expanded and coalesced. Over drusen (n = 45, all calcified), RPE was continuous and thin, photoreceptors were short or absent, and initially intact ELM descended where RPE was absent. In persistent hyperFAF within atrophy and in the OJZ, the RPE was continuous and dysmorphic, photoreceptors were present and short, and BLamD was thick. In the IJZ, mottled FAF corresponded to dissociated RPE atop persistent BLamD. Overall linear growth rate (0.198 mm/ year) typified multifocal GA.

Conclusions: FAF in GA is locally multifactorial, with photoreceptor shortening potentially promoting hyperFAF by increasing incoming excitation light available to RPE fluorophores. RPE dysmorphia may lead to either longer or shorter pathlength for excitation light. At both atrophy initiation and expansion Müller glia are major participants.

Age-related macular degeneration (AMD) causes legal blindness in older adults worldwide and is treated in advanced stages of exudative neovascularization and geographic atrophy (GA). GA is characterized by a degeneration of photoreceptors and their supporting tissues (retinal pigment epithelium [RPE], Bruch membrane [BrM], and choriocapillaris [ChC]), in the setting of characteristic extracellular deposits.^1–3 Slowing the enlargement of existing GA, as seen in 488 nm fundus autofluorescence (blue FAF), was the clinical trial end point for 2023 approvals by the US Food and Drug Administration for inhibitors of complement cascade proteins.^4,5 Clinicopathologic correlation, that is, histology of eyes imaged during life, can both validate diagnostic technology with anatomic ground truth and probe disease pathogenesis. New data as we provide can thus aid patient selection for treatment and interpretation of trial results. 

The principal signal source for FAF is RPE lipofuscin and melanolipofuscin (L/ML). These long-lasting organelles of lysosomal origin accumulate starting prenatally⁶ secondarily to diurnal outer segment disk renewal and in topographic relation to the distribution of photoreceptors.^7,8 Use of FAF in trial end points was supported by detailed observational studies of GA natural history.^9–11 This work established a profound loss of FAF signal in atrophy, distinctive patterns of hyperautofluorescence (hyperFAF) surrounding atrophy, the potential for quantifying atrophy automatically due to high contrast, and slower growth of atrophy toward the fovea than away from it. Integration with cross-sectional imaging from optical coherence tomography (OCT)^12,13 allowed investigation of how FAF signal may be affected by non-RPE tissue layers. 

An early idea that hyperFAF in GA signifies RPE L/ML accumulation to a cytotoxic concentration¹⁴ has been superseded by the discovery of cellular-level explanations for hyperFAF that can be correlated to OCT. Quantitative multimodal microscopy of donor eyes with GA revealed that rounding, stacking, or anterior migration of individual RPE can focally raise FAF signal by lengthening the pathlength of incoming excitation light through RPE fluorophores.^15,16 High-resolution histology demonstrated in the neurosensory retina a precise border of atrophy (external limiting membrane descent to BrM [ELMd]). At the ELMd, normally vertical Müller glia extended horizontally and scrolled in advance of the expanding atrophic area.^16,17 Approaching the ELMd from unaffected retina, the percentage of abnormal RPE morphologies increased and the number of photoreceptors decreased, in a cascading process.¹⁶ This sequence conformed with transitions between in-layer RPE and intraretinal pigmented cells corresponding to hyper-reflective foci of OCT.^18,19 Considering other layers, clinicopathologic correlation showed that a distinctive hyperFAF associated with an “exudative floodplain” in neovascular AMD involved absence of light-absorbing photoreceptors and presence of continuous dysmorphic RPE.^20,21 HypoFAF atrophic spots associated with individual drusen were found to be preceded by stages of hyperFAF involving loss of photoreceptors.²² Accordingly, the onset of clinically visible hyperFAF in GA eyes includes disrupted ellipsoid zone (EZ) and ELM over a thickened RPE-basal lamina (BL)-BrM band.²³ 

Herein, we provide high-resolution histology for an eye with bilateral multifocal GA, foveal sparing, and a relatively thick choroid that was imaged with 485 to 580 nm excitation FAF,²⁴ over a 5-year period. Our imaging hypothesis is that RPE is the principal FAF signal source and other retinal layers, especially overlying photoreceptors, impact signal detected at the camera by modulating light availability to the RPE through absorption.^25,26 We use as a landmark the ELMd to divide a hyperFAF outer junctional zone (OJZ) from a hypoFAF inner junctional zone (IJZ) with puncta of residual FAF.²⁷ The IJZ is at the perimeter of the deeply hypoFAF atrophic area. Building on our FAF staging system for drusen-driven atrophy,²² we assess the integrity of ELM, photoreceptors, and RPE over drusen. 

Methods

Compliance

Retrospective review of medical records and imaging data and the histopathologic study were approved by institutional review boards of the Manhattan Eye, Ear, and Throat Hospital/Northwell Health and the University of Alabama at Birmingham, respectively. This study was conducted in accordance with the Declaration of Helsinki and the Health Insurance Portability and Accountability Act of 1996. 

Clinical Course

An 82-year-old Caucasian woman was monitored over a 5-year period in a private retina practice for bilateral GA secondary to AMD. Ocular history included bilateral cataract extraction with intraocular lens implantation 6 years before presentation, at which time refractive error was −0.25 (OD) and +0.50 (OS). Medical history included cardiovascular disease with mitral valve prolapse, systemic hypertension, and hypercholesterolemia. The patient never smoked. There was no indication of medications for cancer. 

Upon initial presentation, 11 years before death, best-corrected visual acuity (BCVA) was 20/80 and 20/40 in the right and left eyes, respectively. Clinical evaluation and retinal imaging revealed central and peripheral drusen and patchy areas of atrophy in both eyes consistent with a diagnosis of advanced AMD. Over subsequent visits, the atrophy progressed in both eyes. Some soft drusen spontaneously regressed. Superior to the right optic disc, an area of exudative type 1 macular neovascularization was found (and later confirmed in histology). No subretinal hemorrhage was documented, and the exudation spontaneously resolved. At the last registered clinical evaluation, 5 years after presentation and 6 years before death, BCVA was 20/400 in the right eye and 20/100 in the left eye. The patient died of breast cancer at age 93 years. 

Ophthalmoscopic examination and multimodal imaging included OCT (Spectralis, Heidelberg Engineering; automated real time averaging 8–12, quality 32–47), FAF (535–580 excitation, flood-illuminated Topcon TRC-50XF camera), and red-free and near-infrared reflectance. Together these showed central and peripheral large drusen, many refractile drusen, subretinal drusenoid deposits, and fovea-sparing multifocal GA in both eyes. Detailed OCT B-scans of the left eye showed outer retinal and RPE atrophy, large and calcified drusen, and outer retinal tubulation (ORT). 

Histology Overview

Both eyes were recovered 5 hours after death by personnel of the Eye-Bank for Sight Restoration (NY) and opened anteriorly to allow preservative entry.²⁸ The left (index) eye was immersed in phosphate buffered 2.5% glutaraldehyde and 1% paraformaldehyde suitable for epoxy resin high-resolution histology and transmission electron microscopy. The right eye was immersed in phosphate buffered 2.5% glutaraldehyde and 1% paraformaldehyde suitable for immunohistochemical detection of key marker molecules. The current report on correlates of FAF imaging is the first of several describing specific biologic aspects of GA. Forthcoming studies will address vascular pathology, glial response, and fine structure of the OJZ. 

Histology Preparation and Analysis of the Index Case

An ex vivo OCT volume²⁸ for the left eye was tracked to the last clinical volume (Supplementary Video S1). During the 6 years before death, atrophy progressed in the manner seen during the in vivo observation period, seen best in the video with ex vivo near-infrared FAF. In vivo and ex vivo color photography, near-infrared autofluorescence, and near-infrared reflectance were aligned via eye-tracked OCT to subsequent histology (see Supplementary Video S1, 13:81 seconds). Tissues prepared for high-resolution epoxy resin histology allowed visualization of individual RPE granules by light microscopy (see Supplementary Materials). Histology images were manually matched to clinical OCT scans by comparing overall tissue contours, on paper. 

Semi-Quantitative Review of the Index Case

Scans of one section (of five) per each of 140 glass slides 60 µm apart were reviewed by an experienced observer (author C.A.C.) for comparison to clinical FAF and OCT images. A point-to-point correspondence of specific features such as individual drusen in clinical imaging and microscopy is ideal²⁹ but was impeded by several factors in this case. These include a large GA lesion that expanded for 6 years before death, partial detachment of drusen-bearing areas outside the main atrophic area which was adherent, and difficulty in matching stepped histology sections to the 31 clinical B-scans. For these reasons, we opted to compare characteristic FAF patterns (Fig. 1) to characteristic histology patterns. A magnified FAF image from the last clinic visit (see Fig. 1) shows drusen and a main atrophic area. The latter was lined externally by a hyperFAF OJZ and internally by an IJZ with mottled FAF. Within the main area, peninsulas of persistent hyperFAF connected to non-atrophic tissue; islands of persistent hyperFAF did not. 

Figure 1.

 

Representative patterns of fundus autofluorescence. Magnified fundus autofluorescence from the last clinic visit 6 years before death shows characteristic features. The green lines indicate the areas for which comparable histology was explored, as shown in subsequent figures. Drusen exhibit defined stages of FAF^2,22 (Figs. 2, 7). Peninsulas and islands of persistent hyperFAF in the atrophic area are illustrated in Figure 7. The hyperFAF outer junctional zone (OJZ), hypoFAF inner junctional zone (IJZ) with puncta of FAF signal, and hypoFAF atrophic area are illustrated in Figure 8. For comparison, iso FAF uninvolved retina is found in the upper right corner.

We compared layers that could impact FAF signal associated with drusen (RPE, photoreceptors, and basal laminar deposit [BLamD]). Our framework for drusen-driven atrophy (Fig. 2) is based on clinicopathologic correlations that documented hyperFAF preceding hypoFAF.^2,22 Four stages start with drusen visible in color only and end with total absence of photoreceptors and RPE at hypoFAF stage 4. As photoreceptors shorten and disappear (first OS, and then IS), the ELM approaches the druse apex (stage 2). When photoreceptors are absent, an ELM formed by reactive Müller glia alone skims the druse apex, and the RPE layer forms a gap beneath it (stage 3). The ELM descends as a circle partway down the druse slope (stage 3′),² then further down to the druse base (stage 4). 

Figure 2.

 

Stages of drusen-driven atrophy in fundus autofluorescence (FAF). Schematic is derived from previous clinicopathologic correlation (stages 1–4²²; inserted step 3’²) Top row: Affected retinal layers: + = present; – = decreased or not present; –/+ = mixed effects (see Methods and Table 1). Panels a and b refer to an FAF annulus and center, respectively. Middle row: FAF patterns (stage 1 = visible in color only; stage 2 = uniform hyperFAF; stage 3 = annulus of hyperFAF (a) around a center of hypoFAF (b); stage 3’ = larger hypoFAF center; and stage 4 = uniform hypoFAF). Bottom row: One RPE-capped druse and overlying external limiting membrane (ELM, dashed line) are shown. The color scale indicating progressive replacement of lipoprotein-rich soft druse material with hydroxyapatite nodules is qualitative. As photoreceptors shorten and disappear (first OS, and then IS), the ELM approaches the druse apex (stage 2). When photoreceptors are absent, the ELM is formed by reactive Müller glia alone and skims the druse ape. The RPE layer begins to disappear beneath it (stage 3). The ELM forms a circle partway down the druse slope coincident with RPE absence (stage 3’), then further down to the druse base (stage 4). Basal laminar deposit (BLamD) is present at all stages. IS, inner segment; OS, outer segment; RPE, retinal pigment epithelium.

Drusen were analyzed if the overlying ELM was intact or descended as far as the druse base (stage 4; see Fig. 2). Drusen within the main atrophic area were all denuded of RPE, without nearby ELM, and were thus excluded from analysis. By simultaneously viewing scans of 4 to 6 60-µm-stepped sections at approximately 16X on a monitor (using FIJI, https://imagej.net/ij/),³⁰ we identified 45 individual drusen across multiple sections. Of these, 17 were attached, 7 were opposed (detached with the druse close to an indentation in the bacillary layer), 15 were detached (and not close enough to the detached retina to match the indentation), and 6 were mixed (opposed-detached, opposed-attached in different sections). On each section through each druse, 6 layers were scored on a 3 point scale: ELM (1 = intact, 2 = not intact, any size gap, and 3 = at druse base), inner segment (1 = normal, 2 = short, and 3 = absent), outer segment (1 = normal, 2 = short, and 3 = absent), RPE (1 = covering druse, 2 = not intact, any size gap, and 3 = absent), BLamD (1 = absent/ patchy, 2 = thin, and 3 = thick/ late), druse calcification (1 = <1/3 of cross-sectional area, between 1/3 and 2/3, and >2/3). 

We hypothesized that persistent hyperFAF regions (see Fig. 1) represented areas of OJZ that were trapped by the growth of atrophic spots on either side and was thus bounded by ELMd. These were too small to follow over stepped sections to determine their morphology (i.e. extended for peninsula versus focal for islands) using the multi-section viewing method described for drusen. Thus, cross-sectional lengths along BrM are reported for both together. Finally, confluent atrophy areas (>1000 µm wide) bounded by ELMd were assessed for the presence of zones of FAF shown in Figure 1. 

Measurement of Atrophy Area and Growth Rate in Fundus Autofluorescence

To measure the hypoFAF regions corresponding to GA, clinical FAF images were processed by a custom FIJI plugin (Measure_GA, available at https://sites.imagej.net/CreativeComputation/; from within ImageJ see CreativeComputation on the list of sources on the Update menu.). Phansalkar local thresholding with a radius of approximately 0.250 mm was used to binarize the image. An experienced observer (author C.A.C.) drew a polygon loosely surrounding the region judged to be GA, separating the GA from other dark regions, such as vasculature or the optic nerve head. These regions included the confluent main area and small surrounding spots corresponding to stage 4 FAF over drusen, as described above. Within that polygon, dark pixels were counted to compute “area.” Dark pixels which had an (8-connected) neighboring light pixel were counted to compute “perimeter”. The area was scaled to mm², using a universal conversion of 0.288 mm/degrees. Effective radius growth rate was calculated as described³¹ for the 58-month observation period. One time point (11 months post-baseline) was omitted, because it was clearly brighter than the others and had a lower measured area. 

Results

Clinical Imaging

Figure 3 shows multimodal imaging at the last clinic visit, 6 years before death. Color photography (see Fig. 3A) shows sharply demarcated depigmentation superior temporal to the fovea, representing a merging of patches seen earlier (Supplementary Fig. S1). It also reveals refractile drusen throughout the central area, symmetrically distributed around the fovea, with typical soft drusen superior and temporal to the fovea (see Supplementary Fig. S1). By near-infrared reflectance (Fig. 3B), many drusen have punctate reflectivity, and subretinal drusenoid deposits are observable (see Figs. 3B, 3D). Atrophy is visible superior to the fovea but not elsewhere, because the thick choroid in this eye reduces reflections from the sclera.³² By OCT and histology (see Figs. 3D, 3E), the choroid is thick nasally, due to abnormal dilation of the outer choroid and attenuation of the inner choroid. Temporally it is thin overall. Over a 58-month period, GA expanded from 7.8 to 19.5 mm², for an effective radius growth rate of 0.198 mm/ year. 

Figure 3.

 

Multimodal imaging at the last clinic visit, 6 years before death. (A, B, C) Green lines indicate level of the optical coherence tomography (OCT) B-scan (D) and tissue section (E). (A) Color photography shows multifocal areas of hypopigmented geographic atrophy with foveal sparing. The entire central area is replete with refractile drusen with soft drusen superior temporally. (B) By near-infrared reflectance, many drusen have punctate reflectivity. Atrophy is most visible superior to the fovea. A thick choroid overall reduces signal from the sclera. The green arrowhead, subretinal drusenoid deposit; these are also apparent at the top of the panel. (C) Fundus autofluorescence at 435 to 580 nm excitation shows multifocal atrophy that is confluent superior, nasal, and temporal to the fovea, which is spared. (D, E) By OCT (D) and histology (E), choroid (yellow arrowheads) is thick nasally and thin temporally. The pink arrowhead indicates a calcified druse with punctate and linear reflectivity (D) and retina detached from a large druse (E) that is not visible in this section.

FAF shows clearly the continuity between individual drusen and spots of multifocal atrophy. These images are presented as six time points, separately (Fig. 4), and registered and annotated (Supplementary Video S2, Fig. 5). Confluent hypoFAF superior, nasal, and temporal to the fovea is surrounded by smaller hypoFAF spots. The fovea is spared during the observation period but is atrophic by the time of death and subsequent histology. Supplementary Video S2 and Figure 5 show individual drusen progressing through FAF stages 2 to 4, exhibiting different progression dynamics. For example, compare the relatively stable drusen in the superior-temporal near-periphery (orange arrowheads) to others reaching stage 4 (black arrowheads). Two drusen have distinctly heterogenous FAF patterns (teal arrowheads, see Figs. 5C–F). 

Figure 4.

 

Progression of atrophy in 535 to 580 nm excitation fundus autofluorescence. (A–F), Images from six clinic visits are shown, with time before death in years. Images are aligned for continuous viewing in Supplementary Video S2. Selected images are magnified in Figure 4. Many individual spots of atrophy grow and coalesce over this period, beginning with superior-nasal and inferior to the fovea. As these spots expand, islands and peninsulas of hyperFAF may persist in the atrophic area for several years. A hyperFAF border is apparent first nasally, then throughout, by the end of this period. Original circular images were cropped to ovals showing identical regions.

Figure 5.

 

Fundus autofluorescence (FAF) showing stages over drusen. A progression through stages over drusen are shown in Figure 2. Time points in years before the patient’s death are indicated. Variations in illumination cause changes in the FAF signal and the appearance of atrophic areas. Scale bar = 200 µm (A–E). A melt through these images is shown in Supplementary Video S2. (A) Multifocal geographic atrophy with foveal sparing and predominantly stages 4 and 2 FAF over drusen in the central area and near-periphery, as defined.⁸ (B) Geographic atrophy progresses toward the fovea, with coalescence in the nasal, superior, and inferior sectors. FAF over drusen is stage 3 to 3’. (C) As atrophy expands, islands and a peninsula emerge, with FAF over drusen progressing (black arrowhead). FAF remains stable in some regions (yellow/orange arrowheads). (D) Teal arrowheads show the transition between stages 2 and 3 FAF over drusen (teal dotted box magnified in F). Superior and inferior regions reach stages 3 (brown arrowheads) and 4 (black arrowheads). (E) Most FAF over drusen progresses to stage 4 (black arrowheads), except for a stable patch (orange arrowheads). (F) Magnified inset of D, highlights 2 drusen transitioning from stage 2 to 3 (upper teal arrowhead) and stage 3 to 3’ (lower teal arrowhead, scale bar = 100 µm).

Throughout, the main atrophy area is bounded by a hyperFAF OJZ, which may merge with drusen-associated hyperFAF. As atrophic spots expand, they trap bits of former OJZ, creating peninsulas and islands of persistent hyperFAF within the main atrophy area (see Figs. 5C, 5D, respectively). Over time, peninsulas and islands decrease in size and signal intensity more slowly than the junctional zones. Finally, in the hypoFAF IJZ, FAF signal is mottled, with puncta that are visible in magnified, high-quality images (see Figs. 5D, 5E). Thus, the IJZ represents a trailing edge of the overall hypoFAF atrophy. 

Histologic Analysis, Overview

Histologic review of 140 of the 60-µm-stepped sections included 110 through the main atrophic area. Median section length examined was 11.7 mm. Median lengths of atrophic and non-atrophic areas was 4.9 mm and 7.3 mm, respectively. Median druse diameter estimated from the number of stepped sections was 180 µm (Q1 = 120 and Q3 = 300). 

All 45 assessed drusen contained calcific nodules (Table 1), a common end-stage of classic soft (lipoprotein-rich) drusen.³³ We assume that calcified drusen all started as soft drusen. Native soft drusen were not seen, either in the atrophic zone or outside it, nor was basal linear deposit, a thin layer of the same material. Noteworthy incidental findings about drusen include multiple empty lines consistent with cholesterol clefts in 24% of assessed drusen (Supplementary Fig. S2A). Within one partially calcified druse, parallel strands of a blue staining material (same as collagen-containing BLamD and sclera) conformed with “avascular fibrosis” (J.P. Sarks, personal communication July 12, 2024³⁴; see Supplementary Fig. S2B). 

Table 1.

 

Semi-Quantitative Analysis of Drusen and Outer Retinal Layers

 

Drusen Assessment

Figure 6 shows five histologic sections through one druse complex with attached retina that illustrate several principles of FAF variation. The complex is bi-lobed, likely due to confluence of two smaller drusen. It does not have an ELM circle on a single dome, as implied by Figure 2. Rather it has two spots lacking RPE and photoreceptors, each bounded by an ELMd (see Figs. 6A, 6B, 6D, 6E). Between them is a small area with photoreceptors, complete coverage by continuous but dysmorphic (mostly thinned) RPE, and an intact ELM (see Fig. 6C). Across the druse, the Henle fiber/outer nuclear layer is dyslaminate, that is, discrete layers are no longer discernible due to photoreceptor loss and inward translocation of remaining cell bodies. Interleaved Müller glia are stained pale. 

Figure 6.

 

Cross-sectional histology of a large druse complex and overlying retinal layers. Five histological sections (60 µm apart from superior to inferior) through one druse complex. This complex, located inferior temporal to the fovea, has two atrophic areas bounded by external limiting membrane descents (ELMd) (A, B, D, E) and an area with photoreceptors and intact ELM (C). (A–E) The Henle fiber layer/ outer nuclear layer (HFL/ ONL) exhibits dyslamination (ONL depopulation, inward migration of photoreceptor cell bodies across the HFL, and gliosis of interleaved Müller cell processes which are pale-stained. A thick basal laminar deposit (BLamD) encapsulates the complex. Druse content of numerous calcific nodules is dislodged in processing. (A) ELMd are at the druse base, and RPE is absent (stage 4). The orange arrowhead shows most visible lumen of a diminished choriocapillaris. (B) At the druse apex, RPE absence is encircled by an ELMd (stage 3’). The druse base shape verticalizes the RPE layer, for longer pathlength of incoming excitation light (arrow). (C) The druse is encapsulated by dysmorphic RPE (stages 2 and 3). The ELM does not descend but skims over the druse top, because photoreceptors are present. Sloughed RPE cells (arrow) increase the vertical pathlength for excitation light. (D) A short region of RPE absence underlies a shorter region of ELM absence (stages 2 and 3). The druse base shape verticalizes the RPE layer (arrow). (E) Regions of absent RPE and ELM are larger than in D. BrM, Bruch membrane; Ch, choroid; INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segments; OPL, outer plexiform layer; OS, outer segments; RPE, retinal pigment epithelium; SDD, subretinal drusenoid deposits.

Overall, RPE over drusen tended to be thin (see Supplementary Fig. S2A, Fig. 6). Vertical superimposition of cells that might increase the pathlength of incoming excitation light through fluorophores is shown in Figure 6C (sloughed RPE). In Figures 6B and 6E, the RPE layer at the druse base is almost vertical and parallel with incoming light. It was difficult to find hyper-reflective foci by OCT that might indicate intraretinal migration from the apices of drusen. Thus, in this case, at this stage of disease, the contribution of vertical superimposition effects to drusen-associated hyperFAF seems limited to the druse itself. 

Semi-quantitative assessments of contributors to FAF variation over drusen are summarized in Table 1. The RPE was continuous although dysmorphic in 83% of locations. The ELM was continuous in 81% of the locations. Photoreceptor outer segments were degenerate in 97% of locations (absent = 76% and short = 21%) as were inner segments (absent = 40% and short = 57%). Thus, there is a striking disparity between the relative integrity of RPE and ELM and the disintegrity of the inner and outer segments. Nearly all drusen had BLamD (grade 2 [42%] or grade 3 [57%]). In addition to being thick, BLamD was often scalloped in morphology with spherical inward protuberances covered by RPE. Although potentially a signal source for FAF, BLamD cannot account for FAF variability across individual drusen, because of its relatively uniform thickness and persistence. 

Peninsulas and Islands

Figure 7 shows representative persistent hyperFAF (124 islands/peninsulas on 61 slides, median length = 175 µm, minimum and maximum = 43 and 717). These areas had a continuous and overall thick RPE layer, some ectopic pigmented cells internal to the continuous layer, and thin continuous BLamD. Thus, they are distinguishable from ORT, which lacks RPE.³⁵ Lengths of overlying outer segments could not be measured, due to compaction (bending) and presence of debris between them and RPE. Yet outer segment abundance was clearly reduced relative to uninvolved retina temporal to the atrophic area (not shown). The ELM was intact and separated islands and peninsulas from surrounding atrophy. The atrophic area was replete with calcified drusen lacking RPE but still covered by persistent BLamD (variably thin and continuous in 56% of examined sections with atrophy). Cellular processes (presumed Müller glia) interrupted persistent BLamD in others. 

Figure 7.

 

Islands and peninsulas of persistent hyperautofluorescence in atrophy. In an area of persistent hyperFAF, RPE (orange) is continuous, overall thick, with some ectopic pigmented cells, and a thin continuous basal laminar deposit. The presence of RPE distinguishes this formation from the outer retinal tubulation. Overlying photoreceptors have outer segments. Whether debris between outer segments and RPE contains subretinal drusenoid deposits or shed inner segments 111 awaits ultrastructural investigation.¹¹¹ The ELM is intact (green arrowheads) and separates this region from atrophy at the right and left. Atrophic area at right has dissociated RPE atop persistent basal laminar deposit and a druse from which calcified nodules were dislodged. Ch, choroid; HFL/ONL, Henle fiber layer and outer nuclear layer combined due to dyslamination; INL, inner nuclear layer; IPL, inner plexiform layer; OPL, outer plexiform layer; RPE-BL-BrM, RPE-BLamD-Bruch membrane complex; Sc, sclera, v, vein. The arrow indicates most visible lumen of a diminished choriocapillaris.

OJZ, IJZ, and Atrophy

Figure 8A shows an OCT B-scan through the border region (OJZ, ELMd, and IJZ) that is comparable to histological sections (Figs. 8B, 8C). By OCT a pair of ELMd delimit the main atrophic area as well as two deposits outside this area. Within the atrophic area, OCT also reveals calcified drusen, possibly with BLamD caps, and subsidence of OPL. A comparable transition through OJZ, ELMd, and IJZ is shown in histological sections at medium and high magnifications in Figures 8B and 8C, respectively. 

Figure 8.

 

Histology of fundus autofluorescence patterns at the border of atrophy. (A) Optical coherence tomography (OCT) B-scan through drusen-driven atrophy contextualizes histology of junctional zones. This image was captured at the last clinic visit (ART 12, Quality 37) and expanded axially to reveal detail. It is located approximately 1620 µm superior to where histology is taken 6 years later but is nevertheless representative because the atrophy expanded inferiorly. Six green arrowheads indicate descents of the external limiting membrane (ELM). A central area of atrophy is delimited by green arrowheads 3 and 4. Within that area, the OPL subsidence occurs at the pink arrowheads 1 and 2. To the right of pink arrowhead 2, where the OPL is still visible, are junctional zones comparable to the histology in panels B and C. Reflective “crownlike elevations”¹² in the atrophic area are calcified drusen with persistent overlying basal laminar deposit (BLamD). Paired green arrowheads 1 and 2 and 5 and 6 indicate small atrophic spots outside the central area which show ELM descents over individual deposits. The ELM is barely visible over one retinal pigment epithelium (RPE)-capped druse at the far right (no arrowheads). (B) The ELM descent divides a hyper fundus autofluorescence (FAF) outer junctional zone on the right, from the inner junctional zone of atrophy, where hypoFAF is mottled due to autofluorescent puncta (see Figure 1). The remainder of the atrophic zone, at the left, is hypoFAF. The yellow frame indicates the area magnified in panel C. d, druse; green arrowheads, ELM; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; HFL/ONL, Henle fiber layer and outer nuclear layer combined due to dyslamination; RPE-BL-BrM, RPE-BLamD-Bruch membrane complex; Ch, choroid. C. HyperFAF, photoreceptors lacking outer segments, continuous dysmorphic RPE, thick BLamD; Mottled FAF, absent photoreceptors, dissociated RPE (orange arrowhead), thin BLamD; HypoFAF, absent photoreceptors, absent RPE, thin BLamD. Yellow arrowhead, horizontally oriented processes of Müller glia in a subretinal membrane. White arrowhead, BLamD protuberances, attached to (left) and detached from (right) continuous BLamD; White arrow indicates most visible lumen of a diminished choriocapillaris.

By histology, the hyperFAF OJZ contains photoreceptors lacking outer segments, continuous but dysmorphic and often thick RPE, and thick BLamD, some in spheroids of similar size as single RPE cells (see Fig. 8C orange arrowhead). A similar OJZ was visible at 96% of the nasal and 84% of the temporal ELMd. The hypoFAF IJZ exhibits absent photoreceptors and scattered rounded cells containing numerous typical RPE organelles (“dissociated RPE”³⁶), still attached to persistent BLamD. An IJZ with dissociated RPE was visible at 56% of the nasal and 72% of the temporal ELMd. 

Further into the atrophic zone, where the FAF signal is lower than in the IJZ, RPE is absent. A few cone photoreceptors remain in degenerate ORT, that is, lacking inner and outer segments with some cell bodies still present.³⁵ A large ORT (spanning 840 µm) is present under the previously spared fovea. A thin BLamD persists and undulates above BrM, in one of two configurations: over the top of drusen as they calcify into end-stage lesions (see Fig. 6, “tombstones”) or elevated by masses of Müller glial processes (see Fig. 8C). Horizontally oriented processes of Müller glia in subretinal membranes are evident (see Fig. 8C, yellow arrowhead). The neurosensory retina is denuded of photoreceptors, bounded by the ELMd, and is firmly adhered to structures below, either BLamD or BrM. Retinal adhesion to RPE is variable elsewhere (se Fig. 3E). 

Discussion

As a metabolic imaging technique, FAF is advantageous for detecting, quantifying, and monitoring outer retinal atrophy, especially when anchored by eye tracked OCT.³⁷ Expansion of atrophy in FAF is the anatomic end point for trials of recently approved complement inhibitors.^4,5 Existing regions of GA enlarge via interactions that are mediated by cells at the margins of atrophy. Thus, understanding the OJZ is important for validating GA enlargement as an end point and for maximizing the mechanistic insight available from clinical trial imaging datasets. Herein, we elucidate cellular contributors to FAF signal at the initiation, continuation, and denouement of atrophy, respectively, atop drusen and in the OJZ and IJZ. This progression sequence can be glimpsed in previous morphologic and molecular analyses of GA (Table 2). 

Table 2.

 

Morphologic and Molecular Correlates of Fundus Autofluorescence Zones

 

This case of bilateral multifocal GA shared features with many published descriptions. The pattern of atrophy was multifocal, which is seen in 55% of GA eyes.³⁸ Multifocal atrophy starts atop individual drusen and grows faster than unifocal lesions, in proportion to the total perimeter.³¹ Our measured yearly equivalent radius growth rate of 0.198 mm linear is within the error bounds (0.199 ± 0.012 mm) for multifocal GA in a large meta-analysis.³¹ Atrophy growth in our case was both appositional, that is, in continuity with previous atrophy, and appearing de novo, at a degree of separation.³⁸ Foveal sparing,^39–41 seen in approximately two-thirds of GA-affected eyes,^42,43 was apparent as drusen-driven atrophy expanded more toward the near periphery (>3 mm eccentricity) than toward the fovea. At the OJZ, banded/ diffuse hyperFAF like our case is one of several FAF patterns associated with faster growth.^9,44 Our case had venous congestion contributing to a thick choroid, unusual because atrophy is typically associated with a thin choroid. However, a thick choroid with atrophy is common in Asian populations^45,46 and can also occur in persons of European descent.^32,47 We hypothesize that enlarged inner choroid vessels compress the ChC from below. We did not formally analyze ChC for this report but rarefaction was clear by histology (see Figs. 6–8). The history of cardiovascular disease including valve dysfunction, together with the presence of subretinal drusenoid deposits evokes current clinical interest in this association and vascular origins of AMD.^48,49 Our planned histologic studies will be pertinent to this discussion. 

Our study confirms that hypoFAF atrophy not only lacks RPE-derived FAF signal but also represents advanced disease. Atrophy as classically^50,51 and recently described (see Table 2) harbors severe photoreceptor depletion, glia-driven remodeling at ORT, tombstone deposits (calcified drusen and persistent BLamD), and retinal adhesion to underlying tissue.^33,35 BLamD is collagenous but this potential FAF signal source was not apparent in this eye and imaging system. Nevertheless, weak fluorophores in BLamD and possibly in reactive Müller glia continue to generate detectable fluorescence lifetimes that are uncovered in the absence of RPE.⁵² Although some photoreceptors with outer segments are present in ORT, they are unlikely to contribute to FAF signals,⁵³ and they cannot sustain useful vision long-term.⁵⁴ In contrast to BLamD's persistence, all native soft drusen material disappeared or was replaced by calcific nodules. A “fuzzy border” of GA on OCT⁵⁵ is sharp by histology, and it is the ELMd, a bending of the ELM toward BrM in the absence of photoreceptors.¹⁶ This critical landmark is visible on high-quality OCT (quality ≥24 dB in Spectralis) and not on color or FAF.⁵⁶ 

Atop calcified drusen, both the ELM and a layer of dysmorphic RPE were continuous, whereas photoreceptors were short or absent (see Table 1). Gradual loss of photoreceptors accompanied by a circle of ELM sliding down each druse (see Figs. 2, 6) could be visualized as incremental stages of homogenous then annular hyperFAF, before final hypoFAF (see Fig. 4). These drusen are at a later stage of disease than those with hyperpigmentation recently investigated with fluorescence lifetime imaging ophthalmoscopy.⁵⁷ Our multilayer approach to interpreting FAF signal comes from a previous histologic analysis of the outer retinal neurovascular unit (outer plexiform layer to ChC).¹⁷ This analysis revealed gradual deterioration of BrM and ChC, crossing from the non-atrophic to the atrophic area, across the ELMd. Over decades, this deterioration culminates in high-risk drusen material accumulating in the sub-RPE-BL space, with the overlying photoreceptors driven to hypoxia and metabolic insufficiency. This slow process in BrM-ChC contrasted with a relatively precipitous transition (months to years) involving Müller glia, as encapsulated by the ELMd. We reasoned that responding glia reach some threshold of stress at the neurosensory retina, plausibly the focal degeneration of photoreceptors. Bound by the ELMd, the RPE layer can no longer seal a gap. Once the ELM descends to the base of drusen (see Fig. 6A), mechanisms that propel expansion into non-drusen-bearing regions are less clear but certainly involve continued reactivity of Müller glia. Thus normal glial homeostatic functions are likely lost or drastically reduced. 

The OJZ and islands/ peninsulas that started as OJZ exhibit two major contributors to a hyperFAF signal. First, the RPE is continuous, dysmorphic, packed with organelles, and frequently thickened, creating longer pathlength for excitation light. Second, photoreceptor disintegrity, starting with outer segment shortening and eventual disappearance, leads to more incoming light reaching the RPE. Waveguiding may still be possible,²⁵ to up to a point. The hyperFAF OJZ of clinical research has been defined as 200 to 500 µm wide^58–62 with OCT characteristics of RPE thickening and interruption of photoreceptor attributable bands.^23,63,64 These correspond to reduced visual sensitivity via micro-perimetry,^59,60,65 especially directly adjacent to atrophy.⁶⁰ By OCT a large area of ellipsoid zone disruption,⁶⁶ ELM loss,⁶⁷ or reduced “photoreceptor thickness” (top of EZ band to top of RPE-BL-BrM band)⁶⁸ surrounds the atrophic area and may have predictive value for future GA progression.⁶⁹ 

By magnified FAF, an IJZ was defined within the perimeter of the atrophic area and observed to contain scattered punctate FAF on a hypoFAF background. By histology, the IJZ contained scattered nucleated cells and cellular fragments, both full of typical RPE organelles at the same qualitative packing density as cells in the continuous layer. We consider the isolated cells to be largely RPE-originated, and their disposition in the IJZ as evidence for breakup of the RPE layer as Müller glia scroll the remaining photoreceptors at the ELMd. This interpretation is based on our prior studies, starting with a survey of 53 advanced AMD eyes using the same techniques used herein to reveal typical RPE organelles.^36,70 We defined RPE morphologic phenotypes that included stages of anterior migration into the retina and apoptotic shedding of non-nucleated cellular fragments, among others.⁷¹ Phenotypes were assembled into a hypothetical progression sequence that matched well the behavior of hyperreflective foci over drusenoid pigment epithelial detachments.⁷² Separate studies showed that all dysmorphic RPE were undergoing molecular transdifferentiation. Specifically, they lost retinoid markers typical of RPE (RPE65 and CRALBP) and gained immune markers typical of macrophages and microglia (CD68 and CD163),¹⁸ simultaneously gaining longer fluorescence lifetimes by microscopy.¹⁹ Out-of-layer immunoreactive cells were visible as hyper-reflective foci by ex vivo OCT, but also occurred in-layer, consistent with RPE origin. We speculated that this overall transdifferentiation represented RPE epithelial-mesenchyme transition. One set of pigmented, transdifferentiated cells (“dissociated”) was scattered across atrophic areas like those in the IJZ of the index case. 

A possibility that cannot be excluded by the current data is that melanin-bearing cells in the IJZ are phagocytes that ingested RPE and retained autofluorescent organelles. Two author groups independently reported CD68+ cells in the atrophic area or within the neurosensory retina and concluded that they were of myeloid or microglia origin.^73,74 These investigations were driven by immunology questions. Neither performed experiments to exclude RPE origin, for example, by combining selective marker stains with correlative techniques for definitive visualization of RPE organelles.^73–76 The spatiotemporal and microscopic view of an IJZ in the index eye raises the bar for experimentally proving that cells other than RPE account for punctate FAF signal within atrophy. A recent rat model with a true ELMd may be useful for this purpose.⁷⁷ 

The index eye was imaged using a broad excitation spectrum (485–580 nm), which provides signals overlapping those in widely used 488 nm excitation. By using only one excitation system, we could not detect progression-related shifts among fluorophore emissions. We used a camera originally designed to visualize RPE L/ML while avoiding properties of the aged crystalline lens,²⁴ that is, intrinsic autofluorescence and absorption of incoming light due to lens brunescence. With additional absorption by macular xanthophyll pigments,^78,79 excitation at 488 nm has limitations for detecting foveal involvement in GA⁸⁰ and may have reduced image contrast overall. Measurements of GA area with excitation wavelengths of 518 nm and 532 nm and various proprietary image analysis algorithms can be reliable and highly correlated yet still exhibit a systematic offset from each other.^80,81 Thus, any one study or trial must use one imaging regimen from start to completion, as done herein. 

Our analysis strengthens OCT as an anchoring modality for AMD imaging³⁷ by identifying FAF-relevant details in high-quality OCT (see Fig. 8A). We emphasize that visibility of the ELMd depends on image quality (signal-to-noise ratio), which was ≥32 in this case. This information is reported by the Spectralis review software as Q, separately from automated real time averaging (ART). Our past studies indicated the ELM descent was visible at Q ≥24 and not visible at Q = 11 to 19, whereas ART ≥7 was adequate.^56,82 If one does not recognize a defined border in the neurosensory retina (ELMd) that cleanly divides an area of salvageable photoreceptors in the OJZ versus non-salvageable photoreceptors in the atrophic area, one may overestimate how much retina is available for therapeutic rescue and thus pursue treatments that may prove futile. 

By fortifying a histologic basis of clinical FAF variation, our data add to an overall repositioning of L/ML within AMD pathophysiology as benign agents. Previous studies demonstrated loss rather than gain of histological AF and FAF early in disease^83–86 and far greater abundance of known bisretinoid fluorophores in the peripheral than in the central retina.^87–89 A calculation of the photoreceptor to RPE ratio within the central retina shows that the degradative load leading to L/ML does not spatially parallel the risk for AMD onset and early progression.^8,86,90 L/ML-attributable FAF increases 3-fold before age 30 and 2-fold to age 60 years,^91,92 suggesting that L/ML accumulation and accompanying FAF signal increase may represent a mark of maturation rather than aging. 

Strengths of this study are one case with extensive state-of-the-art multimodal imaging and high-quality OCT, a clear view of the fundus in a patient with pseudophakia, broad excitation spectrum FAF to reveal foveal sparing, an estimate of atrophy growth rate, relatively short postmortem delay to fixation and extensively attached neurosensory retina, histology, and microscopy designed for image validation, and a multidisciplinary clinical and laboratory author team. Supportive histology is possible for AMD imaging due to a high prevalence at older ages, and, in this case, an advanced directive registry that allowed patients to pledge future eye donation. Limitations of this study are use of an FAF imaging system not typically used for quantification, atrophy measurements via custom-written software lacking external validation, lack of genotyping, a 6-year gap between the last clinic examination and recovery for histology, the loss of many calcific nodules in tissue processing, and the semi-quantitative nature of histologic assessments. We caution that one case alone cannot elucidate the range of biologic variability, and generalization to other patients should be cautious. The commonality of our findings with previous literature on GA may inform hypothesis-driven research with larger clinical samples, which we encourage. 

In conclusion, we report the first direct clinicopathologic correlation of FAF appearance associated with the anatomic structures used for registration trials in non-neovascular AMD. FAF variation in GA of AMD is multifactorial, with both RPE and photoreceptor morphology contributing at any one locus. Photoreceptor disintegrity with concomitant reduction of incoming light absorption is a major contributor to hyperFAF over drusen. Atrophy initiation at the focal degeneration of photoreceptors over drusen and atrophy expansion both involve Müller glia, and both are visible in FAF. Linking high-quality FAF to OCT is key to future insight, as much detail from high-resolution histology can be seen in living people with current technology. Finally, as atrophy begins over drusen, for which a viable biogenesis model exists,⁸ targeting aspects of drusen biology for therapeutic and preventative measures before atrophy should be prioritized. 

Acknowledgments

The authors thank Yonejung Yoon, MSc, PhD, of The Eye-Bank for Sight Restoration (NYC) for timely retrieval of donor eyes. 

Supported by The Macula Foundation, Inc., New York, NY; unrestricted funds to the Department of Ophthalmology and Visual Sciences (UAB) from Research to Prevent Blindness, Inc., and EyeSight Foundation of Alabama. A.B. reports grants from the Werner Jackstädt Foundation. The Carl G. and Pauline Buck Trust funded the slide scanner. Sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. 

Fundamental research in fundus autofluorescence validation (C.A.C. and J.D.M.) was supported by NIH grants R01EY06109 and R01EY027948. 

Author Contributions: All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Bijon and Freund had full access to all clinical data. Curcio and Messinger had full access to all laboratory data. These authors are responsible for data integrity and accuracy. Study conception and design: K.B.F., C.A.C., M.M.E., J.D.M., A.B., K.R.S., D.S.M., and J.B. Acquisition of data: J.B., J.D.M., A.B., D.S.M., K.B.F., and C.A.C. Analysis and interpretation of data: C.A.C., J.D.M., A.B., K.R.S., D.S.M., M.M.E., J.B., and K.B.F. Writing of the manuscript: C.A.C., J.D.M., A.B., K.R.S., D.S.M., M.M.E., J.B., and K.B.F. 

Meeting Presentation: Portions of this research have been presented at the Association for Research for Vision and Ophthalmology 2024 Annual Meeting (Seattle, WA, USA). 

Disclosure: C.A. Curcio, Heidelberg Engineering (F), Genentech/Hoffman LaRoche (C), Apellis (C), Astellas (C), Boehringer Ingelheim (C), Character Biosciences (C), Osanni (C), Annexon (C), Mobius (C), Ripple (C) (outside this project); J.D. Messinger, None; A. Berlin, None; K.R. Sloan, None; D.S. McLeod, None; M.M. Edwards, None; J. Bijon, None; K.B. Freund, AstraZeneca (C), Apellis Pharmaceuticals (C), EyePoint Pharmaceuticals (C), Genentech (C), Novartis Pharma AG (C), Regeneron Pharmaceuticals, Inc. (C) 

Supplementary Material

Supplementary Video S1. Using a combination of eye tracked in vivo and ex vivo OCT volumes and registration of major retinal vessels in en face imaging technologies, it was possible to approximately align histologic sections with clinical B-scans. The 787 nm autofluorescence obtained ex vivo demonstrates a pattern of atrophy like that. 

Supplementary Video S2. Fundus autofluorescence images from clinical visits 11 years to 6 years before death were registered to show clearly the appearance, growth, and coalescence of atrophic spots associated with drusen.