Müller Cell Changes and Subretinal Membrane Formation in an Eye With Multifocal Geographic Atrophy

Abstract

Purpose

Müller cell morphology and markers were investigated using histology and immunohistochemistry in an eye with clinically documented multifocal geographic atrophy (GA) and correlated with clinical images.

Methods

The donor was followed clinically for 5 years, 6 years before death. The superior posterior pole retina of the right eye was dissected and immunolabeled with antibodies against glial fibrillary acidic protein (GFAP; activated Müller cells and astrocytes) and glutamine synthetase (GS; Müller cells) and Ulex europaeus Agglutinin-1 lectin (blood vessels) before embedding for JB-4 cross section analysis. The inferior macula was cryopreserved. Cryosections were immunolabeled with Müller cell homeostatic and activation markers. Transmission electron microscopy (TEM) of the left eye was used to study ultrastructure changes.

Results

Gross examination demonstrated mottled retinal pigment epithelium (RPE) over presumably calcified drusen. In the macular area, Müller cell processes surrounding both drusen and outer retinal pigmented lesions created a large subretinal membrane. Cryosection analysis demonstrated persistence of aquaporin 4 and GS in Müller cells with both proteins prominently expressed in the subretinal membrane. Increased MC S100B and GFAP expression were also observed in the atrophic area as well as the outer junctional zone. Cryosection labeling and TEM confirmed Müller cell encasing calcified drusen and RPE debris as well as invading basal laminar deposits.

Conclusions

This multifocal GA case demonstrates how MC activation and structural changes surrounding individual drusen could coalesce, contributing to photoreceptor loss. Müller cells penetrating basal laminar deposits and encasing calcified drusen suggests attempting clearing and/or protecting the retina from harmful contents.

Age-related macular degeneration (AMD) causes irreversible central vision loss among aged adults worldwide.¹ Dry AMD, the more prevalent form, results in focal loss of photoreceptors, retinal pigment epithelial (RPE) cells, and choriocapillaris (CC) in the setting of characteristic extracellular deposits. Advanced dry AMD, or geographic atrophy (GA), can present as either unifocal (one large area of atrophy), or multifocal (several areas of atrophy that coalesce).

Beginning early in AMD progression, Müller cells, the primary glial cell in the retina, become activated overlying drusen, as demonstrated by glial fibrillary acidic protein (GFAP) expression.^2–6 As the disease progresses, these glial cells become increasingly activated and remodel, changing shape and location within the retina. Müller cells play a key role in the border of atrophy in the neurosensory retina, recently defined as the descent of the external limiting membrane (ELM) toward Bruch's membrane.^7–10 This border is visible in high quality optical coherence tomography (OCT) and clearly separates an area of photoreceptor depletion in the atrophic area (inner junctional zone [IJZ]) from the surrounding area of potentially salvageable photoreceptors (outer junctional zone [OJZ]). Müller glia at this border are curved rather than vertical. As part of this remodeling process, Müller cells also create subretinal glial membranes in eyes with unifocal GA.^2,4,6,9,10 However, the Müller cell remodeling, and the mechanisms regulating this, in multifocal GA remain poorly defined.

Here, we consider two non-mutually exclusive mechanistic models to explain Müller glial remodeling in multifocal GA. First, we hypothesize that individual drusen act as localized triggers for Müller cell nuclear repositioning and descent of the ELM, generating multiple discrete Müller cell-defined borders that subsequently coalesce into a continuous subretinal glial membrane as atrophy expands. Second, we propose that altered distribution of key Müller cell functional proteins—most notably aquaporin-4 (AQP4) and glutamine synthetase (GS)—reflects a compensatory response to disrupted RPE-mediated metabolic and osmotic homeostasis, rather than a purely degenerative loss of Müller cell function. Consistent with this framework, we recently reported that Müller cells in GA retain expression of essential functional proteins, but with markedly altered cellular domain localization. Most strikingly, the polarized expression of AQP4, a key osmoregulatory protein, shifts drastically in the atrophic area in eyes with unifocal GA.¹¹ Rather than being prominently expressed within endfeet at the inner limiting membrane (ILM), AQP4 is expressed diffusely throughout Müller cell processes with the strongest expression within the subretinal glial membrane. Strong AQP4 expression was also noted within Müller cell processes through all retinal layers at the border of atrophy.

In the present study, we investigated the Müller cell response in an eye with clinically documented multifocal GA using both flatmount and cryosection immunohistochemistry, JB-4 histology, and, in the fellow eye exhibiting similar characteristics, using transmission electron microscopy (TEM). In addition to the expression of signature Müller cell functional proteins, vimentin, cellular retinal binding protein (CRALBP), and GS, we also investigated markers of glial activation, GFAP, and S100B. S100B is a chemokine that is expressed by astrocytes both in the brain and retina as well as some activated Müller cells.^12–14 Although less studied in the retina, S100B has been studied extensively in relation to stroke and Alzheimer's disease where it is believed to contribute to inflammation.^15–18 To our knowledge, this is the first report to demonstrate immunohistochemically the retinal glial changes associated with multifocal GA. A previous report on the fellow eye of this donor addressed histologic correlates of fundus autofluorescence variation, including atrophy initiation at ELM descents over individual drusen.⁷

Methods

Human Donor Information

Retrospective review of medical records, imaging data, and the histopathologic study were approved by institutional review boards of the Manhattan Eye, Ear, and Throat Hospital/Northwell Health, the University of Alabama at Birmingham (UAB), and Johns Hopkins University (JHU) School of Medicine. This study was conducted in accordance with the Declaration of Helsinki and the Health Insurance Portability and Accountability Act of 1996. Utilization of this human tissue was in accordance with approval of the Joint Committee on Clinical Investigation at JHU. The donor presented herein was a 93-year-old Caucasian female patient with clinically diagnosed multifocal GA. She was non-smoker with significant history of cardiovascular disease, including mitral valve prolapse, systemic hypertension, and hypercholesterolemia. Age-matched controls (n = 5) with no ocular abnormalities were used to demonstrate the normal expression patterns of Müller cell proteins. The average age of these control donors was 85 years (±4.3 years). The Müller cell protein expression in three of these control donors was previously reported.¹¹ Table 1 summarizes donor information.

Table 1.

Human Donor Eyes

Patients	Age, Y	Sex	AMD Status	Cause of Death
GA	93	F	Multifocal GA	Breast cancer
Control 1	83	M	Normal	Renal failure
Control 2	80	F	Normal	Cerebrovascular accident; myocardial infarct
Control 3	89	F	Normal	Alzheimer's disease
Control 4	81	M	Normal	Parkinson's disease
Control 5	91	M	Normal	Rectal sheath hematoma

F, female; M, male.

Donor Eye Collection and Tissue Preparation

The GA donor eye was collected 5 hours postmortem by personnel of the Eye-Bank for Sight Restoration (NY) and opened anteriorly to improve fixative penetration.¹⁹ The right eye (OD) was fixed in 4% paraformaldehyde (PFA) in phosphate buffer overnight and then placed in 1% PFA at 4°C for 6 months until shipping and processing. The left eye (OS) was preserved in 1% PFA and 2.5% glutaraldehyde and processed for ultrastructural analysis by light microscopy and TEM. The posterior pole of the OD eye was imaged upon arrival at the Wilmer Eye Institute. The central retina was bisected, with the inferior half cryopreserved as a full-thickness eyecup and the superior half processed as separate retinal and choroidal flatmounts. The choroid is reported separately.²⁰ Control eyes were either fixed exactly as the right eye of the GA donor (N = 2), at 6 hours postortem, or fixed in 2% PFA in phosphate buffered saline (PBS) containing 5% sucrose within 24 hours postmortem (N = 3).

Flatmount Immunohistochemistry

Retinal tissue containing the macula and optic nerve head was processed for flatmount immunohistochemistry as described^2,3 and compared with control posterior pole retinas. Briefly, the retina was blocked in 5% goat serum in tris-buffered saline containing Triton X-100 (TBS-T) and 1% bovine serum albumin (BSA) overnight at 4°C. After washing, the retina was incubated in primary antibody cocktail containing chicken anti-GFAP (Invitrogen, 1:500) and mouse anti-GS (Invitrogen, 1:500) prepared in TBS-T with BSA for 72 hours at 4°C. Following three washes, the retinas were incubated in secondary antibody cocktail containing TBS-T, goat anti-chicken Alexa Fluor 647 (Invitrogen, 1:500), goat anti-mouse cy3 (Jackson Immunoresearch, 1:500) and Fluorescein isothiocyanate (FITC)-conjugated Ulex europaeus Agglutinin 1 (UEA-1) lectin (vascular endothelium; Genetex Inc., Irvine, CA, USA; 1:100) for 48 hours. After final washes, retinas were imaged on a Zeiss 710 confocal microscope. All imaging settings (laser power, gain, etc.) were kept the same for control and GA donor eyes. To visualize overall tissue architecture, differential interference contrast (DIC) images were also collected.

JB-4 Embedding and Analysis

After imaging, the GA retina was fixed flat by placing the tissue between 2 layers of a Nitex mesh in a modified Swinney filter holder, flat fixed in 25% Karnovsky's fixative²¹ and embedded in glycol methacrylate (JB-4; Polysciences, Inc., Warrington, PA, USA), as previously described.⁴ Two and a half-µm-thick sections were cut using a dry glass knife on a microtome (Sorvall MT2-B, Norwalk, CT, USA). Sections were transferred to drops of water on glass slides and dried on a hot plate prior to staining with periodic acid/Schiff's (PAS; Sigma-Aldrich, St. Louis, MO, USA) and hematoxylin (Sigma-Aldrich; HHS16). Images were captured on a Zeiss Photomicroscope II (Carl Zeiss, Inc., Oberkochen, Germany) using a Gryphax NAOS 20 Megapixel Full HD USB 3.0 Color Digital Microscope Camera (Jenoptik, Jena, Germany).

Cryosection Immunohistochemistry

Sections (8-µm thick) were air dried and treated in –20°C methanol for 5 minutes. After air drying, the sections were blocked in 10% goat serum prepared in TBS with 0.1% BSA for 1 hour, and then incubated in primary antibody cocktail (Table 2) prepared in TBS with 0.1% BSA for 2 hours. Sections were washed in TBS and incubated in secondary antibody cocktail prepared in TBS containing DAPI for chromatin (Invitrogen; 1:1000) and UEA lectin (Genetex; 1:100) for 30 minutes. Finally, tissue autofluorescence was quenched by treating sections in Sudan black B (1% freshly prepared in 70% ethanol; Electron Microscopy Sciences; Hatfield, PA, USA) for 10 minutes in the dark followed by washing in distilled water. All steps were carried out at room temperature. Images were captured on a Zeiss 710 confocal microscope and processed using Zen software. Images were collected with a 20 × objective with or without tiling. All image settings (laser power and gain) were kept the same for control and GA donor eyes for each antibody combination. DIC images were also collected.

Table 2.

Primary Antibodies Used in This Study

Antibody	Source, Catalogue #	Working Dilution
Ck-α-GFAP	Millipore ab5541	1:500
Ms-α-GS	Millipore MAB302	1:500
Rb-α-Vimentin	Abcam AB45939	1:500
Rb-α-AQP4	Almone Labs, AQP004	1:400
Ms-α-opsin	Millipore 4886	1:100
Ms-α-CRALBP	Invitrogen WK330690	1:500
Rb-α-S100B	Abcam ab52642	1:200
Ulex europaeus agglutinin 1 lectin (UEA lectin)	Genetex GTX01512	1:100

Ms, mouse; Rb, rabbit; Ck, chicken.

Transmission Electron Microscopy

The fellow (left) eye was fixed, processed, and imaged at UAB as described.⁷ Sections from the left eye were prepared for TEM. For correlative high-resolution histology and electron microscopy, a rectangular tissue block (8 × 12-mm wide) containing the fovea and optic nerve was postfixed in 1% osmium - tannic acid - paraphenylenediamine and embedded in epoxy resin (Polybed 812 resin kit; Electron Microscopy Services, Hatfield, PA, USA). Submicrometer sections (approximately 0.8-µm thick) were cut with a histology grade diamond knife (Diatome Histo knife; Electron Microscopy Services) on an ultramicrotome (Ultracut; Leica) and saved at 30- to 60-µm intervals on glass slides. Interleaved slabs approximately 30-µm-thick were saved for TEM. Selected slabs were re-embedded by first flattening them between two microscope slides on a low temperature hot plate. Under stereoscopic magnification, the flattened section was compared to the image of a submicrometer section. A 2-mm-wide region of interest was dissected out with a sharp scalpel (#11; Graham Field Sterile Ophthalmology Micro Scalpels 2979 #45). This region was placed along the vertical surface of an embedding mold (Pelco Flat embedding mold, Ted Pella Catalog #105; Fisher Scientific) and propped up vertically by a new blank resin blank. New resin was added, and the mold was heated at 75°C for 18 hours. Sections were cut at gold thickness on a Leica UC7 ultramicrotome using a Diatome “Ultra” series diamond knife and placed onto carbon coated slotted copper grids (2 × 1 mm) or uncoated 100 mesh hexagonal copper grids (Electron Microscopy Services). Sections were stained with 2% uranyl acetate and post-stained with Reynolds lead citrate. Images were taken on a H7600 Hitachi TEM (Hitachi High-Technologies Corp., Tokyo, Japan) equipped with an Advanced Microscopy Techniques model XR80, 2KCCD model (Woburn, MA, USA).

Results

Clinical History and Fundus Presentation

The donor had been followed clinically over the course of 5 years, ending 6 years prior to death. Ocular history included bilateral cataract removal with lens implantation. Fundus photography (Fig. 1A) demonstrated drusen in the macular and mid-peripheral regions. Fundus autofluorescence revealed atrophic areas with RPE loss (Fig. 1B). A small area with exudative choroidal neovascularization (CNV) was observed superior to the optic disc at baseline (Fig. 1B arrows) was observed without treatment. There was no exudation in the last clinical examination. Multiple large, calcified drusen as well as central outer retinal and RPE atrophy were evident on the OCT B-scan at the last visit (Figs. 1C, 1D). The choroidal vessels also appeared abnormally dilated (Fig. 1D). Gross images taken prior to removing the retina demonstrated the presence of drusen in the affected macula (Figs. 1E, 1F).

Figure 1.

Multimodal clinical and laboratory imaging. The last clinical images of the multifocal GA donor eye were collected 5 years prior to the patient's death. (A) Fundus photography shows drusen in the macula and perimacula. (B) Fundus autofluorescence revealed the atrophic area with RPE loss as well as a small area with exudative choroidal neovascularization (CNV) superior to the optic disc (arrow). (C, D) The last OCT-B scan revealed multiple large, calcified drusen, as well as central outer retinal and RPE atrophy. The choroidal vessels also appeared abnormally dilated. (E, F) Widefield ex vivo imaging with the retina in place demonstrated drusen in the affected macula. Scale bars indicate 200 µM C and D.

Flatmount Analysis Revealed a Large Discontinuous Subretinal Membrane Created by Müller Cells Encasing Drusen Occupying Atrophic Areas Within the Posterior Pole

The control retinas, when viewed from the side of the ELM, were unremarkable at low magnification (Supplementary Fig. S1A). The honeycomb-like pattern created by Müller cell processes at the ELM were evident at high magnification (Supplementary Fig. S1B). In images of the GA eye, the atrophic areas were well demarcated. In these regions, numerous long, thin processes, positive for GS and GFAP, extended along the outer retinal surface (Fig. 2A). These processes were interrupted frequently by calcified drusen, many of which remained adherent to the retina during dissection from the choroid (Figs. 2, 3). These drusen were enveloped by Müller cell processes (Figs. 2B–D). The density and complexity of these glial processes was best demonstrated at higher magnification and could be segregated by z slices at different focal planes (Figs. 3A–F). Multiple Müller cell processes, positive for both GS and GFAP, extended through the ELM to completely envelope some drusen (Fig. 3B). Between drusen, processes from Müller cells extended horizontally along the outer retinal surface (Fig. 3). In the focal plane containing the ELM, a disrupted ELM was evident with prominent GS staining in areas lacking drusen (Fig. 3C). Müller cell processes also appeared to both encircle and invade other drusen often segregating individual calcified nodules (Figs. 3C–E). Müller cell processes extended along the subretinal surface to fill the gaps created by drusen in the ELM focal plane (Fig. 3F). Therefore, the Müller cells appeared to sequester the retina from drusen components.

Figure 2.

Activated Müller cell processes, viewed from underneath the multifocal GA retina, reorganized around drusen. (A) Half of the macular region stained with GFAP (green) and GS (red). A glial membrane was observed on the outer retinal surface (arrow). The glial processes within the membrane were positive for both GFAP and GS. In some areas, Müller cell processes appeared to be surrounding drusen. (B–D) Higher magnification more clearly shows Müller cell processes, positive for both GS and GFAP, extending over areas previously occupied by drusen mechanically lost during dissection (asterisks). Scale bars indicate 1 mm A and 100 µm B to D.

Figure 3.

High magnification Z series of multifocal GA retinal flatmount stained with GFAP and GS reveals Müller cell processes surrounding drusen. (A) DIC imaging of the retinal flatmount shows two drusen in the upper right and left corners. (B) In the focal plane at the level of the outer retina, Müller cell processes positive for GFAP and GS extended horizontally along the outer retinal surface. In the upper left corner, Müller cell processes created a thick band surrounding the drusen. On the upper right corner, processes were covering the druse. (C) In a focal plane closer to the ELM, the disorganized ELM was obvious with intense GS labeling. Processes were present surrounding the drusen on the upper left corner with some appearing to extend into individual nodules within the druse. The druse on the upper right corner contained visible nodules which were each encircled by GFAP and GS-positive processes. (D) In another area, in the focal plane most external to the retina, multiple Müller cell processes, positive for GFAP and GS, surrounded a large druse. On one side, a wall-like structure was created with Müller cell processes extending over the druse. (E) In a focal plane closer to the ELM, Müller cell processes entering the druse were even more evident. (F) Müller cell processes covered almost the entire gap created by the druse at the level of the ELM. The honeycomb-like pattern of the ELM is evident on either side of the druse. Müller cell processes covering the druse extended horizontally across the retinal surface creating swirls. Scale bars indicate 100 µm.

In one area highlighted in Figure 4, an RPE-capped calcified druse remained adherent to the retina and was evident in the gross photograph. This same area was visualized in the flatmount showing GFAP and GS-positive Müller cell processes extending onto the surface to ensheathe the druse (Fig. 4B). JB-4 sections of the same area, stained with PAS and hematoxylin, revealed a thick basal laminar deposit (BLamD) overlying the druse (Figs. 4C–E). Glial processes were observed surrounding the BLamD and the druse (Figs. 4D, 4E). Melanin containing organelles were observed within Müller cell processes adjacent to the drusen and BLamD (Fig. 4E).

Figure 4.

Müller cell association with calcified drusen in en face and cross-sectional views. (A) Gross photographs of the multifocal GA donor eye showed drusen (arrows) surrounded by RPE. (B) The flatmount retina stained with GFAP (green), GS (red), and UEA lectin (blue), and imaged with the ELM en face, shows the same drusen (arrows) which remained attached to the retina during dissection. RPE were nonspecifically stained with UEA lectin. Müller cell processes were observed overlying drusen as well as between drusen. (C–E) JB-4 sections of the flatmount retina were stained with PAS/hematoxylin. The drusen with multiple nodules were clearly visible as was the BLamD overlying the drusen and in adjacent areas (arrows). Müller cell processes were observed surrounding the drusen and were disorganized arrowheads in E. Scale bars indicate: A and B = 0.5 mm; C and D = 100 µm, and E = 50 µm.

Analysis of Cross-Sections Further Demonstrates the Glial Membrane and its Complexity as Well as Müller Cells Ensheathing Drusen

In the control retina, GFAP immunoreactivity was confined to astrocytes in the nerve fiber layer and isolated Müller cell processes (Supplementary Fig. S2). CRALBP labeled the cell bodies and processes of Müller cells as well as RPE cells (Supplementary Fig. S2). Within the atrophic region of the GA eye, CRALBP expression appeared reduced compared to that in the control eye in the inner retina but was still present. It was more prominent external to the outer plexiform layer, where photoreceptors were absent (Figs. 5A, 5B). Müller cell processes, positive for CRALBP and GFAP, were disorganized throughout the retina, having lost their usual linear structure (Figs. 5A–C). As seen in flatmounts, these processes extended through the outer retina, defined herein as external to the outer plexiform layer regardless of whether photoreceptors were present, and directly contacted drusen (Figs. 5A–D). These drusen created mounds, forcing the layers above to adapt to the new topography. Dense bands of GFAP and CRALBP double-positive processes were observed over drusen and on Bruch's membrane. Some drusen were observed in the affected area with photoreceptors and RPE still present (Fig. 5). In these areas, Müller cells created an ELM descent on either side of drusen when RPE were still present between two drusen (Figs. 5A–D, arrowhead). On the non-atrophic aspect of these ELM descents (OJZ), the outer nuclear layer (ONL) had fewer nuclei but similar thickness compared to unaffected areas located remote to the OJZ. Nuclei, which appear to be within GFAP and CRALBP positive Müller cell processes, were observed on the outer retinal surface on both sides of the drusen as well as between two drusen. A dense line of double-positive processes was also observed running perpendicular to Bruch's membrane (Figs. 5E–H). Pigmented cells, presumably of RPE origin but lacking CRALBP immunoreactivity, were noted on the retinal aspect of this glial membrane (Fig. 5). These cells were often observed mounding on top of one another, particularly adjacent to drusen (Figs. 5G–L). In several areas, nuclei were also observed in the inner and outer plexiform layers, suggesting that Müller cells had migrated throughout the retina as would occur with remodeling.

Figure 5.

Gliosis and retinoid markers of retinal cross sections from the multifocal GA donor eye. (A) Within the affected area, Müller cells were positive for both GFAP (green) and CRALBP (red). DAPI (blue) demonstrates presumed repositioning and dislocation of Müller cell nuclei around drusen and in between two druse (asterisks). Müller cells, positive for both GFAP and CRALBP created the ELM descent (arrowheads) on either side of drusen. (A–C) At the border of atrophy (arrow), RPE cells were also positive for CRALBP. (D) DIC imaging demonstrated pigmented cells, presumably RPE, at the base as well as in between the two calcified drusen. (E–H) A dense band of double-positive processes were observed over the larger drusen (asterisk) with a BLamD (arrows). (I–L) In between drusen, pigmented cells (arrowheads) were surrounded by GFAP/CRALBP-positive Müller cell processes arrowheads. Asterisks indicate calcified drusen in all images. All drusen in this image were devoid of RPE. Scale bars indicate 100 µm A to D, and 50 µm E to L. GCL, ganglion cell layer; INL, inner nuclear layer; ONL, outer nuclear layer.

Müller Cells Exhibited Increased S100B and GFAP Expression in the Atrophic Retina

In control retinas and in unaffected regions of the posterior pole within the GA eye, S100B and GFAP immunoreactivity were confined to astrocytes and isolated Müller cell processes (Supplementary Fig. S2). In the affected macular area, however, Müller cells strongly expressed both GFAP and S100B (Figs. 6A–C). Their processes occupied locations both internal and external to persistent BLamD overlying calcified drusen. In some cases, Müller cell processes extended into the druse itself. Throughout the atrophic area, S100B expression within Müller cells was most prominent in the outer retina and in processes extending subretinally (Figs. 6A, 6B). As was observed with other markers, S100B and GFAP-positive cells/processes were located at the base of each druse. Whereas Müller cells normally have a vertical and linear morphology, this usual arrangement was disrupted by the presence of numerous drusen in the affected areas. Some Müller cells and their processes within the subretinal membrane appeared to lie horizontally along a BLamD or directly in contact with Bruch's membrane. Between drusen, mounds of pigmented cells were often observed in the outer retina, surrounded by GFAP and S100B-positive Müller cell processes (Fig. 6).

Figure 6.

Müller cells were activated and enwrapped calcified drusen in the affected area of the GA eye. (A) S100B (red), GFAP (green), and DAPI (blue) staining through the pathologic area in the case eye. S100B and GFAP-positive processes were also observed above and below a BLamD over two apparently calcified drusen. Müller cell processes overlying drusen had projections into the drusen. A very thin layer of GFAP and S100B-positive glial cells was also observed on the Bruch's membrane side of the BLamD. (B) As in the control areas, S100B was prominently expressed by astrocytes, however, strong expression was also observed in Müller cell processes. (C) GFAP was expressed by astrocytes and activated Müller cells. (D) DIC microscopy demonstrated pigmented cells, presumably RPE aggregates, mounding on one another within the retina. Asterisks indicate drusen. Scale bar indicates 100 µm.

Müller Cell Expression of AQP4 Shifted in the Atrophic Area While GS Expression Persisted

In age-matched control retinas, AQP4 was confined to astrocytes and Müller cell endfeet at the ILM as well as perivascular Müller cell processes in the deep capillary plexus (Supplementary Fig. S2). Little, if any, AQP4 is observed at the ELM in control retinas. In the unaffected areas of the GA posterior pole, GS and AQP4 expression patterns were similar to that observed in controls (data not shown). In the non-atrophic aspect of the ELM descent (OJZ), GS immunostaining was more intense in the outer retina, including Henle's fiber layer, compared to the inner retina (Figs. 7A, 7C). Aquaporin 4 was prominently expressed in the inner retina and perivascular Müller cells but was also diffusely expressed in their radial processes throughout the retina (Figs. 7A, 7B). At the ELM descent, AQP4 immunostaining was increased within Müller cell processes throughout all retinal layers (Figs. 7A, 7B; arrow). Within the IJZ, the atrophic aspect of the ELM descent, Müller cells showed intense staining for AQP4 but very low, if any, GS immunoreactivity (Figs. 7A–C arrow). Directly adjacent to this, within the more atrophic region, GS expression was prominent in the outer retina. Müller cells, positive for AQP4 and GS, were overlying and encapsulating calcified drusen, similar to observations made with other markers. The expression of these proteins, however, was distinct around drusen, GS being more pronounced on the Bruch's membrane aspect of drusen, whereas AQP4 was more prominent on the inner aspect (Figs. 7A–C). Particularly strong expression was noted in the perivascular processes (paired arrows in Figs. 7A–D) as well as in the subretinal glial membrane (Figs. 7A, 7B). In the atrophic area, GS was reduced in the inner retina compared with the controls but increased in the outer retina and subretinal membrane (Figs. 7E–G). Consistent with other Müller cell markers, a continuous thin layer of GS and AQP4-positive cell processes with a few scattered nuclei were present underlying the BLamD. A thick banded structure was observed with DIC imaging between the retina and thin glial membrane (Fig. 7H). Pigmented cells were again noted on the inner aspect of the glial membrane. Here, they were apparently encased by AQP4 and GS-positive Müller cell processes.

Figure 7.

Müller cell expression of GS and AQP4 was altered in the affected area, viewed in cross section. (A) GS (green), AQP4 (red), and DAPI (blue) staining at the atrophic border. (B) In the OJZ, AQP4 was localized to astrocytes and Müller cell endfeet in the nerve fiber as well as perivascular processes (paired arrows) with some diffuse staining in the radial processes. In the atrophic area, AQP4 was also observed in the subretinal glial membrane and in the IJZ. (C) GS was observed in Müller cell radial processes on both the inner and outer junctional zones adjacent to the ELM descent (arrow). Reduced GS expression was observed in the outer retina within the OJZ. Some background autofluorescence of the RPE was observed in the 568 wavelength (green channel). (D) DIC imaging showed the drusen which was surrounded by Müller cell processes. (E) In the adjacent atrophic area, the ONL was replaced by a glial membrane. (F) AQP4 was expressed diffusely in Müller cell processes. Particularly intense staining was noted over drusen (arrowhead), in perivascular processes and in the subretinal membrane. (G) GS staining remained in Müller cell processes even in the atrophic area and in the glial membrane. (H) DIC imaging shows drusen and BLamD. Scale bars indicate 100 µm. Paired arrows indicate perivascular processes positive for AQP4. Arrowheads indicate small drusen with thick BLamD. All drusen in the imaged areas were calcified and lacked RPE. Asterisks indicate the glial membrane in E to H.

Vimentin and Opsin Staining Demonstrated the Müller Cell Structural Changes

Vimentin labeled the entire length of Müller cells in controls and non-pathologic regions of the GA retina, whereas opsin was confined to the cone outer segments (Supplementary Fig. S3). At the atrophic border in the GA retina, vimentin-positive Müller cell processes formed the ELM descent (Figs. 8A, 8C). In the OJZ, Müller cell processes within the Henle fiber layer were thicker than normal, indicative of reactive gliosis. In the IJZ and atrophic area, vimentin-positive Müller cell processes were not linear but rather extended through the retina at different angles and into the subretinal space. As noted with other markers, a very thin layer of vimentin-positive processes with a few nuclei was also observed external to the BLamD (Figs. 8A, 8C, 8G). Opsin was observed in the outer segments. Inner segments were shortened or absent. Mis-localized somatic opsin staining was also observed in cells within Henle fiber layer (Fig. 8). Similar to observations with other antibodies, pigmented cells, presumably RPE or debris from fragmented RPE cells, were embedded within the vimentin-positive membrane. These were separated from the BLamD by Müller cell processes. Importantly, these pigmented cells did not express vimentin, which is a marker of epithelial mesenchymal transition (EMT).

Figure 8.

Müller cell structural changes and redistribution of opsin in the outer junctional zone. (A–C) At the border in the GA donor eye, vimentin-positive Müller cells (green) created the ELM descent (arrow). LM opsin-positive inner segments (red) were observed only in the OJZ. Mislocalized somatic opsin was observed within the ONL and Henle's fiber layer adjacent to the ELM descent. (D) DIC imaging revealed BLamD in the atrophic retina as well as pigmented cells within the vimentin-positive glial membrane. (E–H) High magnification images more clearly demonstrated the vimentin-positive ELM (arrowheads) and Müller cells extending into the subretinal space creating the ELM descent. (F) Somatic opsin mislocalization was also better demonstrated at higher magnification. (H) DIC imaging showed RPE intact on the non-atrophic aspect of the ELM descent (arrowheads). A BLamD was also observed in the atrophic area below the glial cells. Pigmented cells were embedded in the vimentin-positive glial membrane. Importantly, these pigmented cells did not express vimentin, which is an EMT marker.

Glial Cell Processes Invade BLamD and Drusen

As mentioned above, in atrophic areas where RPE cells were lost or only isolated RPE cells remained, Müller cell processes ensheathed individual drusen. This was better observed at higher magnification (Figs. 9A–D). Here, BLamD was also evident over calcified drusen. In these cases, Müller cells created two distinct lamellas, one overlying the BLamD and another beneath the drusen, separating these structures from Bruch's membrane. Müller cell processes, positive for GFAP, were observed penetrating the BLamD as well as enveloping individual hydroxyapatite nodules²² within the drusen (Figs. 9A–D, 9E). Toluidine blue staining and TEM of the fellow eye confirmed Müller cell processes surrounding drusen (Figs. 9F–H). In addition, finger-like processes, identified as Müller cells by their dense and elongated filaments, were noted extending from horizontally oriented Müller cells in the subretinal space. These penetrated into BLamD (Figs. 9G, 9H).

Figure 9.

GFAP staining shows Müller cells enveloping drusen. (A–E) A cross section from the GA donor eye stained with GFAP (green) and DAPI (blue) focusing on a large druse with persistent BLamD and lacking RPE. B Two thick lines of GFAP stained processes were observed overlying the drusen. GFAP processes were also observed within the druse, whereas small, finger-like Müller cell processes penetrated the BLamD covering the druse. C DIC imaging revealed the nodules within the druse as well as the BLamD. D Imaging of the staining along with DIC. E Higher magnification more clearly showed the GFAP-positive processes encasing hydroxyapatite nodules of the druse (arrows). (F) Semi-thin sections of the partner eye stained with toluidine blue revealed a BLamD as well as calcified drusen with apparent Müller cell processes. (G) Transmission electron microscopy also demonstrated Müller cell processes overlying BLamD (asterisk) and RPE (arrows). (H) Finger-like processes, resembling those observed with GFAP staining, were noted extending from horizontally oriented Müller cells in the subretinal space with TEM. Scale bars indicate: A to D, F = 50 µm, E = 10 µm, G = 4 µm, and H = 2 µm. MC, Müller cells; BrM, Bruch's membrane.

Müller Cells Over Drusen Demonstrate Features of ELM Descent Formation

In the unaffected macular retina, as in the control, S100B and GFAP were primarily confined to astrocytes. When drusen was present, however, both proteins were also expressed in overlying and neighboring Müller cell radial processes (Figs. 10A–D). Whereas Müller cell processes did not extend beyond the ELM in these areas, isolated nuclei within GFAP and S100B positive cell processes were observed just below the ELM. Numerous nuclei in the INL and ONL were observed in the outer plexiform layer, suggesting their repositioning within the retina. This reduced the outer plexiform layer thickness and created a concave appearance. RPE cells, while dysmorphic, were still present except at the innermost aspect of the druse, where photoreceptor segments were also missing (Fig. 10D). Müller cell remodeling over individual drusen with RPE was also evident with vimentin staining (Figs. 10E, 10G). Vimentin-positive cells also appeared to extend beyond the ELM along the outermost aspect of drusen. Opsin-positive outer segments were also lost at the apex of drusen and reduced along the edges (Figs. 10E, 10F).

Figure 10.

Glial markers in non - atrophic retina with RPE-capped partially calcified drusen. (A–C) Within the macula, but outside the atrophic area, S100B (red) and GFAP (green) were primarily confined to astrocytes. Some background autofluorescence of the RPE was observed in the 568 wavelength (red channel) due to insufficient Sudan black quenching. Both proteins were also expressed in Müller cell processes directly above drusen (asterisk). The Müller cell processes here lost their linear morphology and were instead oriented towards the top of the druse. (D) DIC imaging demonstrated intact RPE which became thin at the inner aspect of the druse while photoreceptor inner and outer segments were absent. (E–G) Vimentin (green) and opsin (red) staining of cryosections from the multifocal GA donor eye demonstrated Müller cell remodeling surrounding drusen. On either side of the druse, vimentin-positive cells were observed extending beyond the ELM (arrowheads). Opsin-positive inner segments were shortened or missing overlying the apex of the druse (asterisk) but some mislocalized opsin was observed within the retina. (H) DIC imaging revealed RPE intact except at the top of the druse. Scale bars indicate 50 µm.

Discussion

Although previous reports have included donor eyes with multifocal GA,^23,24 these studies did not separate unifocal and multifocal GA in their analysis. Therefore, this is the first comprehensive histological report of Müller cell changes in a retina with multifocal GA. As in unifocal GA, we observed an extensive Müller cell subretinal membrane occupying the atrophic areas in this donor eye. Overall, the changes in Müller cell protein expression were similar to those we recently reported in eyes with unifocal GA.¹¹ As will be discussed below, the multi-lobular nature of the atrophy, however, also created some unique features. These could represent differences between these two forms of GA or could be due to different states of severity as there were more surviving RPE present in the multifocal donor eye. The multifocal atrophic lesions in this eye created many distinct individual boundaries between atrophic and non-atrophic areas, including islands and peninsulas trapped between atrophic spots that initiated atop drusen.⁷ These presented numerous areas to observe ELM descents and associated changes in Müller cell activity and protein expression.

The Müller Cells Ensheathing Drusen Create a Subretinal Membrane

With regard to the glial membrane, the multifocal nature of this atrophy created a more disrupted or disjointed membrane than those observed in unifocal GA.^2,4,9,10 Rather than one large confluent membrane, GFAP and GS-positive processes completely ensheathed individual calcified drusen which separated from Bruch's membrane and adhered to the retinal flatmounts. This occurred both in areas with RPE and photoreceptor degeneration and in adjacent areas where these cells were still present. Cross-sectional analysis confirmed that some Müller cell processes penetrated the calcified drusen and/or were overlying BLamD, whereas others extended along the outer retinal surface, separating these deposits from Bruch's membrane.

The complexity of the Müller cells with their extensive intertwined processes made distinguishing individual cells nearly impossible yet confocal microscopy led to insights of cell behavior. Based on the nuclei present and observations of Sox9-positive nuclei in unifocal GA within the subretinal space (unpublished data), we speculate that at least some Müller cells reposition subretinally and lie horizontally along the outer retinal surface, between the drusen and the inner collagenous layer of Bruch's membrane. This was observed in both cryosection immunohistochemistry with multiple Müller cell markers (GS, vimentin, GFAP, and S100B) as well as with TEM analysis. Frequent interruptions created by drusen gave the membrane a less intricate appearance than membranes in unifocal GA, which are denser and multi-layered.^2,4,9,10 In the most severely affected areas, however, the subretinal glial membrane in this donor was also multi-layered. As these areas also had nearly total RPE and photoreceptor loss,⁷ the differences may represent stages of glial membrane formation corresponding to increased disease severity. Overall, more surviving RPE were present in this multifocal GA eye within the affected macular area compared to unifocal GA eyes we have investigated where RPE were completely atrophic in a large continuous area.^2,4 Although we suspect that RPE cells in the multifocal GA eye were unhealthy, as signified by the presence of drusen and the discontinuity of the monolayer, they were still present. Similar observations were made in the fellow eye.⁷ Interestingly, photoreceptors were missing or distorted over drusen even when RPE were present, as quantified in the fellow eye.⁷

One other notable difference between this eye and previously studied eyes with unifocal GA^2,4 was reduced adhesion between the retina and choroid during dissection for flatmount preparation. Although the retina was more adherent to the choroid than control eyes, it was not as strongly adherent as in eyes with unifocal GA. In addition, the retina in cryosections detached from the choroid even in atrophic areas in this multifocal GA eye. By contrast, in eyes with unifocal GA, we have found the retina to be strongly adherent to the choroid in atrophic areas of cryosections.^2,4 We believe this adhesion is due to the Müller cell processes penetrating Bruch's membrane into the inner choroid.^{2,4,7,9,10,25} In this donor, however, Müller cells appeared to extend horizontally along the BLamD or Bruch's membrane rather than penetrating through Bruch's membrane. It is possible that the composition of BLamD and Bruch's membrane create good substrates for lateral extension of Müller processes, so they do not penetrate the choroid. It is also possible that the choroidal structure is different in this donor, preventing Müller cell invasion. Whereas a separate study has been devoted to choroidal pathologic changes,²⁰ it is worth nothing that the inner choroidal stroma (intercapillary pillars) exhibited severe hyalinization in this eye. This could have impeded Müller cell processes from invading the choroidal stroma, contributing to reduced adhesion.

The Müller cell membrane's complexity, both in this donor and as reported in unifocal GA, may create a stronger barrier than is created by tight junctions in the normal ELM. The glial membrane may be protecting the retina from harmful material in the subretinal space. At the same time, however, it could hinder treatment efficacy. This membrane could also impact the flow of nutrients and waste between the retina and choroid. The discontinuous membrane in this donor with multifocal GA may create less of a barrier to small molecules or drug therapies compared with membranes in eyes with unifocal GA. Further research on the permeability of the glial membrane is required to verify these thoughts.

Müller Cell Remodeling and Retinal Homeostasis: Important Considerations for Disease Progression and Treatment Potential

Müller cell remodeling in response to photoreceptor loss has been studied in both animal models and human donor eyes with retinitis pigmentosa as well as AMD.^2,4,26–30 The Müller cell response to photoreceptor and RPE loss, although secondary to the initial insult, likely impacts retinal homeostasis and treatment efficacy. Given the crucial roles Müller cells play in maintaining homeostatic functions, such as ion and pH regulation and osmolarity, we must understand how Müller cell functions are impacted by remodeling and membrane formation in GA. Whereas Müller cells in this multifocal GA eye retained expression of key Müller cell proteins, AQP4, GS, and CRALBP, protein localizations were altered in the atrophic area. Most strikingly, as we recently reported in unifocal GA eyes and a rat model using subretinal sodium iodate to achieve atrophy bounded by ELM descents,¹¹ AQP4 polarity was disrupted in this multifocal GA eye. Whereas perivascular AQP4 expression persisted, staining increased throughout the Müller cell radial processes and was even stronger in the subretinal membrane. We hypothesize that this spatial shift in expression compensates for reduced debris clearance and osmoregulation by unhealthy or lost RPE. Although this shift in AQP4 is likely beneficial for surviving photoreceptors and other neurons, the increased expression throughout Müller cell processes could negatively impact Müller cell osmoregulation within the retina. This could lead to edema within the retina and/or impaired ability of Müller cells to respond to osmotic stress. It is important to note here that astrocytes, that also express AQP4, could migrate into the subretinal space as has been reported in retinal detachment.³¹ We have, in fact, previously observed astrocyte-like cells in TEM of unifocal GA eyes.⁴ Given the overlap in protein expression, distinguishing astrocyte and Müller cells, particularly if they are dividing or newly formed, is difficult. Based on other markers and cellular morphology, however, we believe most of the AQP4 present in the subretinal membrane represents Müller cells.

We also observed a shift in Müller cell GS expression relative to control eyes, that is, higher levels within subretinal Müller cells/processes and lower levels in the nerve fiber layer. GS persistence comports with our observations on unifocal GA and a rat subretinal sodium iodate model, however, but GS expression within the inner retina was lower in this multifocal GA eye.¹¹ Our data on GS differs from a previous report of GS loss in eyes at various AMD stages.²⁷ The previous report, however, used immuno-electron microscopy and a different antibody than reported herein. Therefore, the discrepancy between our studies is likely due to these technical differences. Additionally, the previous report included both neovascular and atrophic AMD at different stages so the differences could also be due to different disease states.

Müller Cell Activation and Increased S100B Expression Could Contribute to Inflammation in AMD

The expression of both GFAP and S100B within Müller cells in the atrophic retina in the OJZ is not surprising as several reports have demonstrated their activation in AMD.^2,4,6,8,32 A recent single cell transcriptomics study, however, did not detect an increase in GFAP expression and suggested no gliosis occurs in AMD.³³ One possible explanation for this discrepancy is that the 6-mm-diameter tissue punches,³⁴ taken for RNA analysis, included an unknown portion of affected area. The gliotic response is strongest in and around the atrophic area. Further, in early or intermediate AMD, an isolated glial response over individual drusen (120–300 µm diameter),⁷ could easily be missed in a large tissue punch. The transcriptomics study also pooled cells from different stages of AMD with presumably different extents of gliotic response and GFAP expression. It is also possible that the control eyes were not true controls for Müller cells. It is common that older donor eyes with a normal appearing fundus have activated Müller cells throughout the retina, or even pre-retinal membranes.³ These may occur due to prior infections, systemic disease, injuries, or tractional disorders, such as epiretinal membranes. Therefore, control GFAP expression in the single cell transcriptomics study may also not be accurate. Our present and previous results clearly demonstrate reactive Müller cells in AMD.

Of particular interest in this study was the focally increased expression of both GFAP and S100B over solitary drusen in the non-atrophic retina. S100B is a chemokine which exerts both intracellular and extracellular functions. We have observed similar S100B increases in eyes with unifocal GA and our rat sodium iodate model so this is not unique to multifocal GA (Edwards, unpublished data). Müller cells and astrocytes can secrete S100B, contributing to increased extracellular concentration. Whereas nanomolar concentrations are neuroprotective, in micromolar concentrations, S100B exerts pro-inflammatory and neurotoxic effects.^13,14 Müller cells may also secrete S100B into the vitreous and aqueous humor, creating a marker of activation. S100B's role as a chemokine has led to extensive studies regarding its role in Alzheimer's disease and other neurodegenerative disorders.^17,35,36 Given the increase noted with isolated drusen, outside the atrophic area, Müller cell S100B expression could contribute to the inflammatory response in AMD. Further research is required to determine when in AMD this increase occurs as well as how this contributes to AMD pathology. One important consideration with regard to S100B is that it could also be secreted by astrocytes or choroidal cells in the affected area and taken up by Müller cells. The fact that we did not observe increased S100B expression in astrocytes, combined with the increased expression of another activation marker, GFAP, supports increased Müller cell expression.

Individual Drusen Provide Clues Regarding ELM Descent and Membrane Formation

Although it seems reasonable to assume that the glial membrane forms by Müller cells extending processes into the subretinal space, individual drusen suggest an additional scenario. Over each druse, presumed Müller cell nuclei were observed repositioning into the ONL and beyond the ELM even in areas where photoreceptors and RPE were present. This suggests that Müller cell bodies reposition in creating the ELM descent and possibly the subretinal membrane, rather than just extending processes from cell bodies remaining in the inner nuclear layer. We previously reported ELM descents over individual drusen in non-atrophic areas and postulated that atrophy begins when Müller glia permanently fill a gap in the RPE monolayer.^7,37 The retinal remodeling over drusen creates an unusual environment where Müller cells and RPE can directly interact, which is not normally the case. How these cells influence one another is not yet understood. Based on our observations, we propose the following scenario for membrane formation. As drusen increase in number and size, Müller cells and processes surrounding these extend horizontally along the subretinal surface toward one another. These individual areas eventually coalesce to create an adhesive membrane, as seen on our flatmount preparation (Fig. 2).

Causes and Consequences of Müller Cell Remodeling Associated With Drusen

In this case, we observed the Müller cells and processes remodeling over isolated drusen even in areas with RPE, but not photoreceptors, present. Here, we observed increased activation, evidenced by GFAP and S100B expression, but Müller cell processes did not penetrate and invade the drusen. Müller cell structural remodeling was more severe in areas with RPE and photoreceptor loss. Because RPE and photoreceptors were most often lost together in this donor, it is difficult to speculate whether it is the loss of RPE or photoreceptors that stimulates the remodeling. Given their close relationship at the ELM, we posit that the photoreceptor loss, or stress, is the primary stimulus, as supported by quantitative analysis in other clinically characterized eyes.^7,37 It is nevertheless possible that Müller cells respond to signals from RPE cells or components of drusen. Future studies will investigate these ideas using additional donor eyes.

The Müller cell remodeling around drusen likely has several impacts. These Müller cells extending onto Bruch's membrane and BLamD potentially reduce the flow of nutrients from the choroid to the retina and RPE in areas already affected by BLamD and attenuated CC.²⁰ As discussed with regard to protein expression, this drastic remodeling of Müller cells likely impacts their ability to support neuronal synapses and retinal homeostasis.^38,39 Such changes would contribute to retinal degeneration in AMD. Future studies will examine other Müller cell metabolic and homeostatic proteins to fully appreciate the changes in these cells associated with drusen. The retina remodeling in GA will also impact treatment efficacy.

The extension of Müller cells into drusen and surrounding individual calcified nodules confirms our previous observation.⁴ We present the hypothesis that Müller cells around drusen are attempting to remove deposits or calcification from drusen. The cells responsible for removal of tombstone deposits like the calcified drusen in this case (some capped by persistent BLamD) have not been previously identified and Müller cells are strong candidates. Much of this cellular detail is visible in OCT, and some have been defined histologically^7,40 but not with cell-type-specific markers as done herein. Alternatively, Müller cells may remodel around drusen to protect the retina from drusen contents. Although this requires further research, Müller cell activation is likely associated with increased cytokine production which could influence inflammation and disease progression in AMD.

Müller Cells and Migrating RPE

Another interesting observation in this case was pigmented cells and debris, of likely RPE origin, that had migrated into the retina seemingly along or within Müller cell processes. These pigmented cells, presumably RPE, were present as far inwardly as the INL. The idea that Müller cells are facilitating RPE migration through the retina toward retinal vessels has been suggested previously.^41–43 Interestingly, these pigmented cells lacked CRALBP expression, supporting the idea that RPE undergo molecular transdifferentiation, possibly EMT in AMD.^44,45 It is important to note, however, that these cells did not express vimentin, as one would expect for cells undergoing EMT. One other possibility is that Müller cells, microglia, or macrophages have digested dead RPE and are channeling them through the retina. Unfortunately, our antibodies for microglia (Iba1) or macrophages (CD68) did not work well on these sections, perhaps due to the length of fixation.

Conclusions

One limitation of the present study is that it is based on a single set of donor eyes. Therefore, one must consider the influence that other health issues could have exerted on the retinal health. Of particular note in this donor was systemic vascular disease, including hypertension and hyperlipidemia, which could have affected both the retinal and choroidal vasculature, contributing to disease progression in this donor. We have reduced the influence of these factors by looking at sections outside the atrophic area, often within the same section in the macula as well as outside the macula, for comparison. Moreover, our results regarding Müller cell remodeling and protein expression changes are consistent with those obtained from eyes with unifocal GA that we have investigated¹¹ (unpublished data). It is also important to note that the Müller cells response in GA is not homogenous as has been reported in other retinal degenerations.^46,47 Multiple Müller cell subpopulations emerge in response to injury. Future studies will be needed to understand these fully. Another limitation of this study is the temporal gap between the last clinical images and the death. Whereas we acknowledge that the atrophy progressed over this time, the progression characteristics were similar over the course of the disease, as supported by a growth rate analysis in the two eyes.⁴⁸ Our data are significant for providing molecular information about the status of cells that are increasingly visible through multimodal OCT-anchored clinical imaging. Thus, a natural history of what we describe in histology may be determined with precision, in vivo, in well-powered clinical imaging studies.