The Onion Sign in neovascular age-related macular degeneration represents cholesterol crystals

Abstract

Purpose

To investigate the frequency, natural evolution and histological correlates of layered, hyperreflective, sub-retinal pigment epithelium (sub-RPE) lines, known as the Onion Sign, in neovascular age-related macular degeneration (nvAMD).

Design

Retrospective observational cohort study; an experimental laboratory study.

Participants

Two hundred thirty eyes of 150 consecutive patients with nvAMD; 40 human donor eyes with clinical and histopathologic diagnosis of nvAMD.

Methods

Spectral-domain optical coherence tomography (SD-OCT), near-infrared reflectance (nIR), color fundus images and medical charts were reviewed. Donor eyes underwent multimodal ex vivo imaging including SD-OCT before processing for high-resolution histology.

Main Outcome Measures

Presence of layered, hyperreflective sub-RPE lines, qualitative analysis of their change in appearance over time with SD-OCT, histological correlates of these lines, and associated findings within surrounding tissues.

Results

Sixteen of 230 eyes of patients (7.0%) and 2 of 40 donor eyes (5.0%) with nvAMD had layered, hyperreflective sub-RPE lines on SD-OCT imaging. These appeared as refractile, yellow-gray exudates on color imaging and hyperreflective lesions on nIR. In all 16 eyes, the Onion Sign persisted in follow-up for up to 5 years, with fluctuations in the abundance of lines and associated with intraretinal hyperreflective foci. Patients with the Onion Sign were disproportionately taking cholesterol-lowering medications (p = 0.025). Histology of 2 donor eyes revealed that hyperreflective lines correlated with clefts created by extraction of cholesterol crystals during tissue processing. Fluid surrounding crystals contained lipid yet was distinct from oily drusen. Intraretinal hyperreflective foci correlated with intraretinal RPE and lipid-filled cells of probable monocyte origin.

Conclusion

Persistent and dynamic, the Onion Sign represents sub-RPE cholesterol crystal precipitation in aqueous environment. The frequency of the Onion Sign in nvAMD in a referral practice and a pathology archive is 5–7%. Associations include use of cholesterol-lowering medication and intraretinal hyperreflective foci attributable to RPE cells and lipid-filled cells of monocyte origin.

Introduction

The Onion Sign was first described by Mukkamala et al as a novel spectral-domain optical coherence tomography (SD-OCT) finding of layered hyperreflective lines in the sub-retinal pigment epithelium (RPE) space, usually associated with chronic exudation from type 1 neovascularization (NV) in patients with age-related macular degeneration (AMD).¹ Typically associated with intense signal upon near-infrared reflectance scanning laser ophthalmoscopy (NIR-SLO), the Onion Sign was proposed by its discoverers to represent layers of precipitated lipid amidst chronic exudation^{1, 2} after also considering collagen or fibrin.¹ Others authors suggested fibrovascular scarring,³ mechanical strain on Bruch’s membrane, and dystrophic calcification in drusen regression^{4, 5} (Supplementary Table, available online) as possible histological correlates.

Independently Christakopoulous et al⁶ hypothesized that hyperreflective lines in the Onion Sign represent cholesterol crystals, because they are transparent and not associated with shadowing, like calcification. This hypothesis is credible, because cholesterol crystals appear as linear, highly reflecting structures in atherosclerotic coronary artery plaques viewed by SD-OCT⁷ in which reflections are generated from interfaces between crystal surfaces and surrounding tissue. This designation for cardiovascular disease was validated in autopsy and endarterial specimens analyzed by ex vivo OCT and subsequent histology^{8, 9}. Tissue processing using ethanol dissolves crystals, leaving distinctive lucent clefts¹⁰. In human eye pathology, clefts have been reported within one disciform scar secondary to AMD¹¹ yet not in aged and early AMD eyes, which have cholesterol-rich lipoprotein deposits in drusen and Bruch’s membrane¹².

Our goals are to document frequency and natural history of the Onion Sign in AMD patients and resolve the controversy behind the hyperreflective material with histology of two cases identified among archival donor eyes by ex vivo SD-OCT. By confirming the cholesterol crystal hypothesis, our data reinforce the potential synergy of in vivo SD-OCT, ex vivo SD-OCT, and histology for molecular discovery. This information is valuable to clinicians as it aids in the understanding of pathophysiology underlying AMD as well as informing diagnosis and therapeutic decision-making.

Methods

The Western Institutional Review Board and the Institutional Review Board at UAB approved this retrospective, observational cohort study and experimental study, respectively. Research complied with the Health Insurance Portability and Accountability Act and adhered to the tenets of the Declaration of Helsinki.

Retrospective observational cohort study

Medical billing records from March 15, 2014 to September 15, 2014 were used to identify consecutive patients seen by a single physician (K.B.F) in a vitreoretinal referral practice located in New York, NY, with a diagnosis code 362.52 for nvAMD in the International Statistical Classification of Diseases and Related Health Problems, Ninth Revision (ICD-9). All patients had received anti-vascular endothelial growth factor (VEGF) therapy within the course of follow-up, although this was not an inclusion requirement. All patient medical charts, color fundus photography obtained with the Topcon Retinal Camera TRC 50DX (Topcon Corporation, Tokyo, Japan), SD-OCT and simultaneous NIR-SLO using a light stimulus of 815 nm obtained with the Spectralis (Heidelberg Engineering, Heidelberg, Germany) were reviewed. The Onion Sign was identified as the presence of layered, hyperreflective sub-RPE bands with SD-OCT imaging and correlated with findings on color fundus photography and NIR-SLO imaging. Serial eye-tracked SD-OCT scans were used to perform qualitative analysis of the Onion Sign from the time of first detection to the most recent visit. The SD-OCT scanning protocol used in all eyes comprised parallel horizontal line scans over the area of interest, each scan spaced approximately 150 to 250 µm apart, with automatic real-time (ART) averaging set between 16 to 32. The abundance of lines within the Onion Sign was taken as increased if the number or the lengths of hyperreflective lines increased, and decreased if numbers or lengths decreased.

To investigate the possibility of an association of the Onion Sign with systemic hypercholesterolemia, the use of an oral cholesterol-lowering medication was recorded. The association of medication use with the occurrence of the Onion Sign assessed for statistical significance using Fisher’s exact test and SPSS software version 22.0 (IBM Corporation, New York, USA). A p-value of less than 0.05 was accepted as significant.

Histopathology study

Neovascular AMD eyes were identified through an ex vivo imaging screen of eyes accessioned for research purposes from non-diabetic white donors to the Alabama Eye Bank during the period 1996–2012. Median death-to-preservation time was 3:49 hours (range, 0:40–11:40 hours). Eyes were preserved by immersion in 1% paraformaldehyde and 2.5% glutaraldehyde in 0.1M phosphate buffer following anterior segment excision. Following removal of vitreous, maculas were photographed in color on a stereomicroscope (SMZ-U, Nikon, Melville NY), using 35 mm color film (1996–2005) or a digital camera (CoolPix, Nikon; 2005–2009). For both image formats, pigmentary changes were accentuated with oblique trans-illumination via a dark-field base, and drusen were accentuated using epi-illumination via a ring light affixed to the objective of the dissecting scope. Eyes underwent additional multimodal ex vivo imaging including SD-OCT when prepared for histology (2011–2013). From each globe, an 8 mm diameter full-thickness tissue punch containing the fovea and temporal portion of the optic nerve head was removed with a trephine. This punch was held in a tissue holder (courtesy J. Fischer, Heidelberg Engineering) mounted on a Spectralis (Supplementary Figure 1, available online). The holder was a closed chamber with a 60-diopter lens in the front (facing into the SD-OCT instrument) and a slot to hold the tissue with the inner limiting membrane facing forward, in the back. Using tissues stabilized in this manner, we performed a 30°×20° SD-OCT volume (143 scans, 30 µm spacing, with ART set at 25) and red-free SLO. Ex vivo SD-OCT differs distinctly from its in vivo counterpart yet is interpretable with experience. Post-mortem edema overall reduces contrast in the neurosensory retina and underlying choroid¹³. The RPE-Bruch’s membrane band is usually visible. The two synaptic layers, photoreceptor ellipsoid zone, and choroidal vessels are also visible in the best-preserved specimens. Fluid is hyporeflective. Eyes also underwent NIR-SLO and autofluorescence SLO with excitation wavelengths of 488 nm for lipofuscin-melanolipofuscin and 787 nm for melanosomes.

Macular tissue punches were post-fixed by osmium tannic acid paraphenylenediamine to accentuate extracellular lipid and embedded in epoxy resin (PolyBed 812, Polysciences, Warrington PA)¹⁴. Sub-micrometer-thick sections were stained with 1% toluidine blue, and reviewed and photodocumented with a 60× oil-immersion objective (numerical aperture = 1.4) and digital camera (XC10, Olympus; 1900× viewing magnification on a monitor). Histologic sections were annotated and uploaded to the Project MACULA online digital microscope (http://projectmacula). Forty eyes of 40 donors (25 women, 86.9 ± 5.9 years; 15 men, 82.2 ± 7.3 years) had nvAMD, defined by fibrovascular scarring in the presence of severe RPE change plus drusen and/or basal linear deposits (BLinD)¹⁵. Two nvAMD eyes exhibiting highly hyperreflective lines in the sub-RPE compartment consistent with the Onion Sign were identified. We use the nomenclature of Staurenghi et al¹⁶ for SD-OCT bands and Zanzottera et al¹⁷ for RPE morphology in histology.

Results

Clinical imaging and associations

A cohort of 230 eyes of 150 consecutive patients (mean age of 84 years, 108 women, 42 men) with nvAMD was identified. Of these, 16 eyes of 15 patients (7.0%; mean age of 82 years, 13 women, 2 men) were found to exhibit the Onion Sign with SD-OCT imaging. The Onion Sign corresponded to refractile yellow-gray exudates on color fundus photography and hyperreflective lesions on NIR-SLO imaging in all 16 eyes (Figure 1).

Figure 1. Multimodal imaging of the Onion Sign in neovascular age-related macular degeneration.

A 72-year-old white woman with exudative changes. A. Color fundus photograph (RGB) shows yellow-gray exudation just nasal to the fovea and circinate yellow exudates. B. NIR-SLO (NIR-refl) shows hyperreflectivity of the yellow-gray exudates nasal to the fovea. C. Early fluorescein angiogram showing hyperfluorescence temporal to the fovea consistent with neovascularization. D. Late fluorescein angiogram showing late leakage, also consistent with neovascularization, but not co-localizing with the Onion Sign. E. SD-OCT scan through the fovea showing the Onion Sign nasal to the fovea, consisting of multiple layered hyperreflective lines below the RPE. There are multiple intraretinal hyperreflective foci temporal to the fovea. F. SD-OCT scan through the fovea 3 years later shows persistence of the Onion Sign, with fewer hyperreflective lines.

Qualitative analysis of serial SD-OCT scans revealed that the Onion Sign persisted in all 16 eyes over the follow-up duration (mean 3.7 years, range 0.5 – 5 years). Fluctuations in the appearance of the Onion Sign were noted, with 13 eyes exhibiting decreased abundance of lines and 3 eyes showing increased abundance of lines over the follow up period. In addition, all 16 eyes with the Onion Sign were found to have adjacent areas of multiple intraretinal hyperreflective foci with SD-OCT imaging (Figure 2).

Figure 2. The Onion Sign in neovascular age-related macular degeneration fluctuates in size and complexity.

A 67-year-old white woman with exudative macular changes. A. Color fundus photography (RGB) shows refractile yellow exudation temporal to the macula which occurred while receiving intravitreal ranibizumab injections. B. Color fundus photograph taken 3 years later shows the same refractile yellow exudation. C,D. NIR-SLO (NIR refl) and SD-OCT performed at this visit show that the hyperreflective lesion on NIR-SLO corresponds to an Onion Sign consisting of multiple layered hyperreflective lines below elevated RPE. E,F. Corresponding NIR-SLO and SD-OCT taken 1 year later show persistence of the Onion Sign, with reduced dome height and fewer clearly discernable hyperreflective lines. G,H. NIR-SLO and SD-OCT through the refractile exudates taken 3 years later show persistence of the Onion Sign, with increased dome height and more hyperreflective lines.

Of the 135 patients without the Onion Sign, 54 patients (40%) reported using oral cholesterol-lowering medication. This rate was consistent with the percentage of United States adults reporting use of these agents (39.1–54.4%¹⁸). Of the 15 patients with the Onion Sign, 11 patients (73%) were using cholesterol-lowering medications. This association was significant (p<0.025).

Histopathology

Of 40 donor eyes with histopathologic diagnosis of nvAMD, 2 eyes (5.0%) exhibited an Onion Sign on ex vivo SD-OCT imaging. Ex vivo SD-OCT and histology for Case 1, a 98-year-old white woman, are shown in Figure 3. Figures 3A,C show temporal to the foveal center a ~1.2 × 0.66 mm dome-shaped RPE elevation with a hyperreflective surface and an outwardly convex base. Figure 3A shows within the dome 3 groups of thin hyperreflective lines, radiating from a point on the temporal side, resembling a clinical Onion Sign. Histology in Figure 3B shows that the dome has 3 groups of linear clefts at angles corresponding exactly to the hyperreflective bands, surrounded by hemorrhage (abundant erythrocytes without formed vascular endothelium) and overlying an outwardly convex Bruch’s membrane. Six additional Onion Signs were visible by ex vivo SD-OCT in this eye, including 2 more domes and 4 low flat elevations. All corresponded to reflective plaques visible in red-free SLO (not shown). In a dense fibrovascular scar with both subretinal and sub-RPE components were several lucent slits consistent with cholesterol clefts (not shown),¹¹ which had not been apparent by prior ex vivo imaging.

Figure 3. Ex vivo SD-OCT and histology of the Onion Sign.

Sub-micrometer epoxy section, toluidine blue; 98-year-old white woman with neovascular AMD. A. Ex vivo SD-OCT shows 3 groups of thin highly hyperreflective lines (arrowheads) within an overall hyperreflective RPE elevation. The foveal center is nasal to the elevation (off the left edge). B. Histology reveals cholesterol clefts clustered in 3 planes corresponding to the hyperreflective lines in panel A (arrowheads). Bruch’s membrane (black arrowheads) is bowed outward. RPE, retinal pigment epithelium; asterisk, basal laminar deposit; Ch, choroid; Sc, sclera; g, Friedman lipid globule⁶¹. Bar, 100 µm. C. Ex vivo SD-OCT through the same elevation at a location inferior to A shows 3 hyperreflective foci in the neurosensory retina that are magnified in D (red), E (green), and F (teal). Bar in E, 25 µm, applies to D–F. D–F. Intraretinal RPE has full complement of black spindle-shaped melanosomes and green-staining lipofuscin granules (D). Cells of presumed monocyte origin with tan-staining lipid droplets and sparse (E) or absent (F) RPE melanosomes.

Figure 3C shows the RPE elevation of Case 1 in a more inferiorly located B-scan, where it is associated with 3 hyperreflective foci within the neurosensory retina, each with distinct cellular content. Figure 3D shows Intraretinal RPE cells, which are irregularly shaped and packed with black spindle-shaped melanosomes and green-staining lipofuscin granules. Figure 3E,F shows cells with lipid droplets stained tan signifying lipid and few (Figure 3E) or absent (Figure 3F) RPE granules. Relative to the intraretinal RPE in Figure 3D, these tan-staining cells are larger and more spherical in shape, and they are clustered. We interpret cells in Figure 3E,F as phagocytes of monocyte origin that ingested and retained varying numbers of RPE granules while also differentiating into foam cells. Elsewhere in these sections are intraretinal fluid pockets and more hyperreflective foci (not shown).

Ex vivo multimodal imaging and histology for Case 2, a 91-year-old white man, are shown in Figures 4 and 5. Figure 4A is a color image showing a large, vertically elongated atrophic central area with punctate hyperpigmentation and ringed 270° by hyperpigmentation, representing an RPE detachment (PED). Within the central 0.7 mm, the PED is visible through the foveal floor as yellowish with punctate pigmentation, surrounded by golden hyperpigmentation with a sharp central border. Figure 4B shows that by 787 nm autofluorescence imaging, the PED is hypoautofluorescent with lobulated hyperautofluorescent borders. Figure 4C,D shows two ex vivo SD-OCT scans through the PED, which measured 3.720 × 2.583 mm (superior-inferior × nasal-temporal). Intraretinal fluid pockets 1.38 mm and 0.69 mm in length were located temporal and nasal to the PED, respectively, and a 0.15 long mm-fluid pocket was present in the fovea (not shown). Increased light transmitted through the PED apex (Figure 4C,D) revealed numerous horizontal hyperreflective lines with scattered tiny hyperreflective spots against an inhomogeneous background reflectivity. One intense reflectivity source appeared as a thin line on SD-OCT and a plaque on NIR-SLO (not shown), suggesting a thin flat shape. The RPE band on the PED dome was irregular, with small hyperreflective spots between RPE and photoreceptors. In two locations on the PED dome (0.150 and 0.300 mm in length), the RPE band was thicker and associated with hyperreflective spots within the overlying neurosensory retina.

Figure 4. Ex vivo imaging and histology of an RPE detachment (PED) with an Onion Sign and associated hyperreflective foci.

Sub-micrometer epoxy section, toluidine blue; 91-year-old white man with neovascular AMD. A. Color photography shows RPE atrophy (yellow) with a central PED (green). Cotton wool spots (white clumps between the PED and the optic nerve head) are visible in color and NIR-SLO (not shown); they are not visible in 787 nm autofluorescence (panel B). B. 787 nm autofluorescence shows that the PED (green) is hypo-autofluorescent with a hyper-autofluorescent border (yellow). C, D. SD-OCT scans through (C) and inferior (D) to the foveal center show horizontal hyperreflective lines within the PED (green). Median length of clefts is 105.1 µm, range 39.9–533.9 µm. Hyperreflective foci F–I are magnified in panels F–I. E. The PED contains numerous lucent needle-shaped clefts that are cholesterol crystals extracted by histological processing. Cells in the fluid (pink) are presumed macrophages. F–I. Presumed macrophages have polydisperse irregular inclusions. Intraretinal RPE have spindle-shaped melanosomes and monodisperse green-stained lipofuscin. Fluid (f). Black arrows, ELM. Insert G was brightened to highlight a hyperreflective spot believed to represent one or several cells.

Figure 5. Contents and microenvironment of a large PED with an Onion Sign.

Same eye as Figure 4. A. Cholesterol clefts (1), and three fluid phases containing lipid. (2–4); #4 denotes flocculent material apparently originating from lipoprotein-derived debris in adjoining basal linear deposit (BLinD). Teal arrow, choriocapillary with clear plasma. Yellow arrows, Bruch’s membrane. B. At the margin of the Onion Sign-bearing PED, fluid meets oily BLinD at a phase boundary (black arrow). Debris adheres to the PED under-surface (black arrowheads). BLamD, basal laminar deposit. Bar, 20 µm.

Histology for Case 2 shows that the PED contains lucent linear clefts (Figures 4E and 5A, #1) cutting across several phases of fluid (Figure 5A, #2–4) that correspond to the hyperreflective lines and background reflectivity in SD-OCT, respectively. The fluid phases contain lipid, indicated by tan staining and contrast with clear plasma in choriocapillaries (Figure 5A, arrow; Figure 5A, #2, #3). Fluid from the PED and oily BLinD flanking the PED meet at a distinct phase boundary, emphasizing their different physical characteristics (Figure 5B, arrow). Judging from debris adhering to the PED under-surface (Figure 5B, arrowheads) and flocculent material within the PED (Figure 5A #4), the fluid may arise from within BLinD and dislodge pieces of this material. Within the PED fluid, macrophages with irregularly shaped cell bodies and inclusions that are large, deep brown, polydisperse, and unevenly packed account for the hyperreflective spots (Figure 4E). Sloughed RPE cells together with small areas of RPE atrophy on the PED apex account for the irregular pigmentation and 787 nm autofluorescence pattern in the en face view (Figure 4A–B). Intraretinal cysts containing tan-staining fluid with lipid droplets and cells shown in Figures 4F–I together account for the intraretinal hyporeflective cavities and hyperreflective spots, respectively. The cells have evenly-spaced spindle-shaped melanosomes and green-to-bronze staining lipofuscin granules consistent with Intraretinal RPE (Figure 4G,H) or polydisperse irregular inclusions consistent with phagocytes of monocyte origin (Figures 4F,I).

Discussion

To our knowledge, this study is the first to report frequency, natural history, and histological correlates of the Onion Sign in neovascular AMD. A long-lasting yet dynamic structure, the Onion Sign was visible in ~5–7% of nvAMD eyes and positively associated with intraretinal/ subretinal hyperreflective foci and intraretinal fluid. A possible association between the Onion Sign and systemic hypercholesterolemia was suggested by a significantly higher use of cholesterol-lowering medications in cases compared to controls. Histology of two cases with RPE elevation and sub-RPE hemorrhage/ fluid revealed that hyperreflective lines correlated with clefts created by the extraction of cholesterol crystals during tissue processing (Supplementary Figure 6, available online). Such crystals were reported in human eye pathology^{11, 19} and in experimental hypercholesterolemia^{20, 21} where birefringence before processing and clefts after processing can be directly demonstrated (Supplementary Figure 6; available online). Our data thus support the original hypotheses that sub-RPE lipids are contributory¹ and that the contributing lipid has the chemical and physical form of crystalline cholesterol⁶.

We believe that the histology cases from these clinically undocumented donors were nvAMD. Case 1 had disciform degeneration in the setting of basal laminar deposit and BLinD. Case 2 had a serous PED, intraretinal cysts, and hyperreflective spots in the same setting, yet a specific NV membrane was not found, likely due to the spacing of our standard section planes. For example, type 3 neovascularization is focal^{22, 23} and could have been missed. However, the combination of a PED and intraretinal fluid in Case 2 merited consideration as nvAMD. Although histologic analysis used eyes different from those imaged clinically, attribution of hyperreflective lines to cholesterol clefts is justified, as follows: highly similar appearance of ex vivo and in vivo hyperreflective lines, distinctive appearance of cholesterol clefts, the clinically observed transmissivity characteristics⁶, histologically validated imaging of crystals in cardiovascular disease, similar clinical and histologic prevalence in nvAMD eyes, and excellent preservation of donor eyes reducing probability of artifact. Because donor tissues were recovered prior to 2006, it can be concluded that the Onion Sign is not a consequence of anti-VEGF therapy, although we cannot exclude an effect of other treatments unknown to us. Eyes untreated by anti-VEGF therapy might be more likely to progress quickly to scars. The Onion Sign may develop preferentially in eyes with chronic exudation with absence of rapid neovascularization growth or hemorrhage from type 1 neovascularization. Chronic treatment with anti-VEGF agents in eyes showing some resistance and persistent leakage may predispose to this finding. Thus the similar frequency in a clinic population and treatment-naïve donors was somewhat surprising. Although we have not observed Onion Signs in non-nvAMD eyes in our clinical practice, others have⁴ (Supplementary Table), making it premature to conclude that this sign localizes exclusively to neovascular disease. It is plausible that Onion Signs occur in serous PEDs and subclinical focal neovascularization indistinguishable by fluorescein or indocyanine green angiography. We restricted our current investigations to nvAMD.

Cholesterol crystal formation is important for atherosclerotic cardiovascular disease, forming early, instigating inflammation, and contributing to sudden cardiac events.²⁴ OCT imaging of coronary artery plaques yields curvilinear hyperreflective lines corresponding to histologically confirmed crystals.⁷ Cholesterol has 2 chemical forms, unesterified (UC) and esterified (EC, bound to a fatty acid). It also has 3 physical forms that vary in proportions of UC, EC, and solubilizing phospholipid: droplets (EC, UC, phospholipid), membranes (UC, phospholipid), and crystals (UC only)²⁵. Crystal formation in super-saturated solutions is affected by local physical factors, with cooler temperatures, lower pH, and higher saturation promoting formation²⁶, and degree of hydration affecting crystal shape (monohydrate, rhomboid plates; anhydrous, aciculate or needle-shaped)²⁷. Our data suggest that appreciable lipid content in fluid (Case 2) and red blood cells (Case 1)²⁸ are potential contributors to UC super-saturation^{24, 29}.

It appears that super-saturation and subsequent crystal formation in PED requires an aqueous microenvironment that is distinct from the oily microenvironment of soft drusen and BLinD³⁰ (Figure 5). Drusen are extracellular lesions in the sub-RPE compartment that provide a cleavage plane for invading neovessels³¹. The largest volumetric component of drusen is lipid especially EC, UC, and phosphatidylcholine, attributed to RPE-secreted apolipoprotein B, E-containing lipoproteins in a pathway driven by dietary lipid recycling in the outer retina³². Evidence for this model includes accumulation in Bruch’s membrane throughout adulthood of solid particles with classic lipoprotein morphology containing cholesteryl esters rich in linoleate, RPE expression of hallmark lipoprotein secretion genes, secretion of apolipoprotein E-immunoreactive lipoprotein-like particles by highly-differentiated cultured RPE, and the inconsistent association of AMD at the person level with plasma lipoprotein levels³³. Another cholesterol source contributing to sub-RPE super-saturation is also lipoproteins of systemic origin. The absence of cholesterol crystals from Bruch’s membrane and drusen has been thought a crucial difference between AMD and systemic atherosclerosis, two diseases unified by early lipoprotein deposition in vessel walls³⁴.. Thus we propose that crystals represented by the Onion Sign have an important systemic contribution via plasma exudates, supported by the association with self-reported cholesterol-lowering medication. Although this retrospective study was limited by the lack of serum cholesterol data, this initial association between the Onion Sign and self-reported use of oral cholesterol-lowering medication can encourage future studies to investigate this relationship in detail. If our clinical findings are replicated, it may be possible that Onion Signs in AMD patients signify systemic hypercholesterolemia that should be monitored.

Sub-RPE cholesterol crystallization joins etiologically diverse retinal disorders with crystals, including Bietti crystalline dystrophy (CYP4V2 mutations)^{19, 35} cerebrotendinous xanthomatosis (CYP27A1 mutations)³⁶, intraretinal crystals in macular telangiectasia type 2³⁷ and in AMD², pre-retinal crystals associated with triamcinolone usage³⁸, and refractile hydroxyapatite spherules in calcifying drusen.^{39, 40} The Onion Sign is also distinct from other sub-RPE hyperreflective material such as fibrovascular scars with horizontal reflectivities caused by densely packed planes of collagen and fibroblasts. Other SD-OCT features with potential as crystalline are “plaques”, single hyperreflective horizontal lines parallel to and very close to Bruch’s membrane, described by Fleckenstein et al⁴¹ and used as progression markers⁴². Yet cholesterol crystals in this location have not been illustrated by major AMD histological surveys.^{12, 43–47} Whether these plaques are Onion precursors thus awaits further information.

Beyond showing cholesterol clefts, the histological cases provided unexpected insight into RPE fate over localized elevations of two different compositions (hemorrhage and fluid). For the first time, hyperreflective foci associated with intraretinal fluid in ex vivo SD-OCT were tracked to specific individual cells. Hyperreflective foci found in all 16 clinical cases are now plausibly also attributable to such cells. Recently we extensively surveyed RPE morphology in advanced AMD, also using high-resolution histology, to disclose characteristic RPE organelles and basal laminar deposit.¹⁷ Two cellular phenotypes with a full complement of RPE granules that apparently form an anterior migratory pathway¹⁷ were also present in our two cases. Many authors consider these cells macrophages or microglia^48–51. In Case 1 some cells conferring hyperreflectivity are more readily attributable to cells of monocyte origin that phagocytized RPE, retaining telltale melanosomes, while also engorging with lipid droplets, like foam cells, which are hyperreflective in coronary artery plaques⁷. Because we have not seen a resident population of non-pigmented cells in the subretinal space⁵² of human donor eyes, we suggest that the lipid-filled cells are microglia that moved from inner to outer retina, as they do in degenerations and injuries⁵³. Intriguingly, mouse microglia in vitro can take up 7-ketocholesterol, a pro-inflammatory oxysterol found in aging Bruch’s membrane and drusen⁵⁴, and form intracellular lipid droplets⁵⁵. It is tempting to speculate that microglia constitute one population of hyperreflective spots and anteriorly migrating RPE another. One way to test this hypothesis is to determine if cells are separable by size, clumping, motility, and characteristics in multimodal imaging studies^{49, 56}.

Although histology of a clinically characterized eye is considered a gold standard, our study notably identified a correlate to an in vivo SD-OCT imaging sign in an ex vivo SD-OCT screen of archived donor eyes. Limitations of this study include absence of crystal quantification in clinical and histological PED and molecular phenotyping of intraretinal hyperreflective cells. Nevertheless, data support the concept that SD-OCT is an essential tool for characterizing donor eyes in the pathology laboratory^57–59. Targeting single cells of known reflectivity characteristics for labeling studies will be a powerful means to linking molecular discovery with contemporary clinical imaging. This capability is important, because the combination of comprehensive anatomic pathology and excellent clinical imaging is treasured in ophthalmology yet still a rarity⁶⁰, and laboratory animals lacking a macula do not yet faithfully replicate AMD.

Abbreviations

AMD

age-related macular degeneration

ART

automatic real-time averaging

BLinD

basal linear deposit

EC

esterified cholesterol

nIR

near infrared reflectance

NV

neovascularization

nvAMD

neovascular AMD

PED

retinal pigment epithelium detachment

RPE

retinal pigment epithelium

SD-OCT

spectral domain optical coherence tomography

SLO

scanning laser ophthalmoscope

UC

unesterified cholesterol

VEGF

vascular endothelial growth factor