Retinal pathological features and proteome signatures of Alzheimer's disease

Abstract

Alzheimer's disease (AD) pathologies were discovered in the accessible neurosensory retina. However, their exact nature and topographical distribution, particularly in the early stages of functional impairment, and how they relate to disease progression in the brain remain largely unknown. To better understand the pathological features of AD in the retina, we conducted an extensive histopathological and biochemical investigation of postmortem retina and brain tissues from 86 human donors. Quantitative examination of superior and inferior temporal retinas from mild cognitive impairment (MCI) and AD patients compared to those with normal cognition (NC) revealed significant increases in amyloid β-protein (Aβ₄₂) forms and novel intraneuronal Aβ oligomers (AβOi), which were closely associated with exacerbated retinal macrogliosis, microgliosis, and tissue atrophy. These pathologies were unevenly distributed across retinal layers and geometrical areas, with the inner layers and peripheral subregions exhibiting most pronounced accumulations in the MCI and AD versus NC retinas. While microgliosis was increased in the retina of these patients, the proportion of microglial cells engaging in Aβ uptake was reduced. Female AD patients exhibited higher levels of retinal microgliosis than males. Notably, retinal Aβ₄₂, S100 calcium-binding protein B⁺ macrogliosis, and atrophy correlated with severity of brain Aβ pathology, tauopathy, and atrophy, and most retinal pathologies reflected Braak staging. All retinal biomarkers correlated with the cognitive scores, with retinal Aβ₄₂, far-peripheral AβOi and microgliosis displaying the strongest correlations. Proteomic analysis of AD retinas revealed activation of specific inflammatory and neurodegenerative processes and inhibition of oxidative phosphorylation/mitochondrial, and photoreceptor-related pathways. This study identifies and maps retinopathy in MCI and AD patients, demonstrating the quantitative relationship with brain pathology and cognition, and may lead to reliable retinal biomarkers for noninvasive retinal screening and monitoring of AD.

Keywords: Eye; GFAP; IBA1; Immune responses; Neurodegenerative disorders; Ocular abnormalities; S100β; scFvA13-intraneuronal oligomers.

Fig. 1

Spatiotemporal distribution of Aβ₄₂ burden in retinas of MCI and AD patients and relations to brain pathology and cognition. a Illustration depicts analyzed retinal cross-sections in predefined geometrical regions including superior- and inferior temporal (ST/IT) strips (orange) extending from the optic disc (OD) to the ora serrata and separated into subregions: central (C), mid-periphery (M) and far periphery (F). Schematic flow-diagram describes human donor eyes and brains allocated for histological and protein analyses (N = subjects). b Fluorescence micrographs of retinal cross-sections from MCI and AD patients compared to normal cognition (NC) controls. Tissues were immunolabeled for GFAP⁺-macroglia (green), IBA1⁺-microglia (red), 12F4⁺-Aβ₄₂ (white), and DAPI⁺-nuclei (blue; dashed lines indicate margins of analyzed layers between the inner and outer limiting membranes–ILM/OLM). Scale bar: 50 µm. Right micrographs are from the same individuals immunolabeled with 12F4⁺-Aβ₄₂ using peroxidase-based 3,3′diaminobenzidine (DAB) and hematoxylin counterstaining. Scale bar: 20 µm. c Violin plots display quantitative-IHC analysis of retinal (r)Aβ₄₂-immunoreactive area in age- and sex-matched patients with premortem clinical diagnoses of NC (n = 17), MCI (n = 10), or AD (n = 18), and paired-brain (b)Aβ-plaque severity scores in NC (n = 6), MCI (n = 10), and AD (n = 17) patients. Red circle represents an ADAD patient with an A260V mutation in presenilin-1 (PSEN1). d Retinal Aβ_1–42 levels determined by ELISA are shown in an additional cohort of NC and AD patients (n = 14; ADAD patient with PSEN1-A431E mutation, red circle). e TEM-micrographs from AD patients’ retina: Left, 12F4⁺-immunogold Aβ₄₂-positive black puncta signals at high-magnification (red arrow) in the ILM/innermost layers. Scale bar: 200 nm. Middle: 3D-reconstruction of vertical/en face TEM images show rAβ₄₂ plaque ultrastructure with fibril arms emanating from its dense core and Aβ-containing deposits (red arrowheads). Scale bar: 1 μm. Right, Aβ₄₂ plaque (black arrow) and deposits within Müller cell (MC) endfeet (red arrows). Scale bar: 0.2 µm. f Pie charts display Aβ₄₂ distribution across the inner retina (IR), outer retina (OR), and C, M, and F subregions: raw data and normalized per retinal thickness (density); higher burden in darker red. g Violin plot displays rAβ₄₂ density for C, M, and F subregions. h Definition of inner retina (IR) and outer retina (OR) in a cross-section. Scale bar: 10 μm. i Aβ₄₂ burden in IR vs. OR; percentages indicate rAβ₄₂ area in IR of total area. Statistics: red or blue asterisks mark significance relative to NC or MCI, respectively. P_d–diagnostic groups; P_r–C, M, vs. F subregions; P_L–IR vs. OR layers; P_i–interactions. j Scatterplot presents correlations between rAβ₄₂ area and Aβ plaques in total brain (gray) or EC (orange). k–l Mid-sagittal brain illustration and heatmap show color-grading magnitude of Pearson’s correlation coefficient (r) values with multivariable Holm-Bonferroni adjusted P- values (asterisks) between rAβ₄₂ burden and brain pathology: Aβ-(P)laques, neuropil threads (NT), and neurofibrillary tangles (NFT) in the hippocampus (Hipp), superior (S.) frontal (F. Ctx) and temporal (temp, T. Ctx) gyrus, S. parietal lobule (P. Ctx), entorhinal (EC), primary visual (PV), and visual association (VA) cortices. m Pearson’s correlation between rAβ₄₂ burden and BRAAK stage. n Subjects were stratified based on high(H) or low(L) brain ATN-histopathology severity and plotted based on rAβ₄₂ burden; extrapolated dotted-gray line marks rAβ₄₂ level separating ATN^H from ATN^L individuals. o Pearson’s correlations between rAβ₄₂ area or bAβ burden and the Mini-Mental State Examination (MMSE)-cognitive scores. Data points are presented with group means ± SEMs. Filled and empty circles represent women and men, respectively. Median and lower and upper quartiles are indicated on each violin plot. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, by one-way or two-way ANOVA and Tukey’s post hoc multiple comparison test, or by two-tailed paired (parenthesis) or unpaired Student’s t test

Fig. 2

Identification and mapping of scFvA13⁺-AβOi in the retina of MCI and AD patients. a Representative microscopic images of postmortem brain cross-sections show the existence of intracellular Aβ oligomers (rAβOi) detected by scFvA13 (red), a conformation-sensitive and sequence-specific antibody in the format of a single-chain Fv fragment (scFv), selectively recognizing AD-relevant AβOs inside cortical pyramidal neurons in AD patients with a minimal signal in NC controls. b Microscopic images of retinal cross-sections showing rAβOi inside βIII-tubulin⁺-ganglion cells and nerve fibers (green) in MCI and AD patients, while less so in NC controls. Scale bar: 15 µm. Insert image: AβOi-positive retinal ganglion cell (RGC). Scale bar: 3 µm. c TEM micrograph shows subcellular localization of 12F4⁺-Aβ₄₂ (DAB, black) in RGC’s endoplasmic reticulum (red arrowhead). Scale bar: 1 µm. d A quantitative analysis of scFvA13⁺-AβOi-IR area in ST/IT retina in MCI and AD patients vs. NC controls (n = 31; red circle, an ADAD patient with PSEN1-A260V mutation). Data presented as median and lower and upper quartiles. e Bar graph with individual data points displays the scFvA13⁺-AβOi-area colocalized within βIII-tubulin⁺-RGCs. f Pie chart of inner retinal AβOi immunoreactive area distributed in C, M, and F subregions; higher burden shown by darker pink. g Quantitative scFvA13⁺-AβOi-IR area for C, M, and F subregions in the ST retina in MCI and AD patients vs. NC controls. Statistics: red asterisks mark significance relative to the NC control group. P_i–interactions; P_d–diagnostic groups; P_r–C, M, vs. F subregions. h–k Pearson’s correlation coefficient (r) analysis between scFvA13⁺-rAβOi load and (h) rAβ₄₂ area, (i) brain (b)Aβ-plaque, (j) bNFT, and (k) bNT severity scores. l. Retinal AβOi-burden in human donors stratified based on high(H) versus low(L) brain ATN histopathological scores; extrapolated dotted-gray line marks the rAβOi level separating ATN^H from ATN^L individuals. m Pearson’s correlations of ST/IT rAβOi vs. MMSE cognitive scores. Data points are presented with group means ± SEMs. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, by 1-way or 2-way ANOVA with Tukey’s post hoc multiple comparison test

Fig. 3

Distribution of macrogliosis in retinas of MCI and AD patients and relations to disease status. a Representative fluorescence micrographs of retinal cross-sections immunolabeled for S100β⁺ (red)- or GFAP⁺ (green), markers of reactive astrocytes and Müller glia. Retinal (r)GFAP⁺ macrogliosis is detected surrounding sites of 12F4⁺-Aβ₄₂ deposits (red), especially in the ganglion cell layer (GCL) in patients with MCI or AD versus NC. White arrows indicate Aβ colocalized within GFAP⁺ macroglia. b, c Violin plots display quantitative IHC analyses of (b) rS100β-immunoreactive areas in ST/IT retina; total n = 20 patients, and (c) rS100β-positive area per C, M, and F subregion. d Representative images from AD patients stained for 4G8⁺-Aβ (red), GFAP⁺ reactive astrocytes (green), and DAPI nuclei (blue). e. A quantitative IHC analysis of GFAP-positive areas in the ST/IT retina in patients with NC (n = 16), MCI (n = 8) and AD (n = 17; red circle, an ADAD patient with PSEN1-A260V mutation). f Pie charts show GFAP⁺ macrogliosis distribution (raw and normalized to tissue thickness) in the inner retina (IR) and outer retina (OR) and in C, M, and F subregions; higher burden shown by darker green. g, h Quantitative GFAP-positive area analyses in patients with MCI (n = 8–9), AD (n = 12–17), and NC (n = 15–16) in the ST/IT retina, separated for C, M, and F subregions (g), and inner versus outer retinal (IR vs. OR) layers (h). GFAP⁺ macrogliosis is almost exclusively detected in the IR layers (% of total area). i–l Pearson’s correlations of rS100β⁺ or rGFAP⁺ macrogliosis against (i) rAβ₄₂-immunoreactive area, (j) rAβOi-immunoreactive area, (k) brain Aβ plaque score, and (l) NFT score. m rS100β⁺ and rGFAP⁺ macrogliosis burden in subjects stratified based on high(H) or low(L) brain ATN histopathological scores; extrapolated dotted-gray lines mark rS100β⁺ level (but not rGFAP⁺-gliosis) separating ATN^H from ATN^L individuals. n. Pearson’s correlations of rS100β⁺ or rGFAP⁺ macrogliosis against MMSE cognitive scores. Data points are presented with group means ± SEMs. Filled and empty circles represent women and men, respectively. Median and lower and upper quartiles are indicated on each violin plot; red asterisks mark significance relative to the NC control group: P_i–interactions, P_r–C, M, vs. F subregions, P_L–retinal IR vs. OR layers, P_d–diagnostic groups. Statistics: *P < 0.05, **P < 0.01, ***P < 0.001, by one-way or two-way ANOVA and Tukey’s post hoc multiple comparison test

Fig. 4

Distribution of retinal microgliosis, Aβ₄₂ phagocytosis, and relationships to disease status. a Representative fluorescence micrographs showing IBA1⁺ microgliosis (red) in retinal cross-sections from NC, MCI, and AD patients. Scale bar: 20 µm. b Violin plot displays quantitative IHC analysis of rIBA1⁺-immunoreactive area in subjects with NC (n = 15), MCI (n = 9), and AD (n = 15; red circle, an ADAD patient with PSEN1-A260V mutation). c Bar graph displays rIBA1⁺ microgliosis by sex in NC (n = 9F/6 M), MCI (n = 6F/3 M,) and AD groups (n = 5F/10 M). d Fluorescence micrograph shows rIBA1⁺ microgliosis (red) colocalized at sites of 12F4⁺-Aβ₄₂ deposits (white) with GFAP⁺ macrogliosis (green) and DAPI nuclei (blue). Scale bar: 20 µm. Retinal IBA1⁺ microglia often internalize rAβ₄₂ (enlarged images). e Quantitative analysis of co-localized 12F4⁺-Aβ₄₂ puncta count with IBA1⁺ microglial cells. f Percent 12F4⁺-Aβ₄₂ puncta count co-localized with IBA1⁺ microglia of total retinal IBA1⁺ microglia. g Pie charts show rIBA1⁺ microgliosis distribution (raw and normalized to tissue thickness) in IR, and OR, and in C, M, and F subregions, with higher burden marked by darker yellow. h, i Quantitative IBA1-positive area in the ST/IT retinas of patients with NC (n = 14–15), MCI (n = 9–10) and AD (n = 14–15), separated for (h) C, M, and F subregions and for (i) inner versus outer retinal (IR vs. OR) layers. j–m Scatterplot displays Pearson’s correlations between rIBA1⁺ microgliosis and (j) rAβ₄₂, (k) rGFAP, (l) brain Aβ plaque, and (m) NFT scores. n rIBA1⁺ microgliosis in subjects stratified based on high(H) or low(L) brain ATN-histopathology; extrapolated dotted-gray line marks potential rIBA1⁺ microgliosis level for separating ATN^H from ATN^L individuals. o Scatterplot displays Pearson’s correlation between rIBA1.⁺ microgliosis and MMSE cognitive scores. Data points are presented with group means ± SEMs. Filled and empty circles represent women and men, respectively. Median and lower and upper quartiles are indicated on each violin plot; red asterisks mark significance relative to the NC control group: P_d–diagnostic groups; P_s–sex groups; P_i–interactions. Statistics: *P < 0.05, **P < 0.01, ***P < 0.001, by one-way or two-way ANOVA and Tukey’s post hoc multiple comparison test, or Student t test (in parenthesis)

Fig. 5

Retinal atrophy in MCI and AD patients in relation to retinal and brain pathologies and cognition. a A reduction in tissue thickness is shown in representative retinal cross-sections from an AD patient (132 µm) vs. NC control (176 µm). Thickness was measured from ILM to OLM (purple dashed lines). Scale bar: 30 µm. b Histomorphometric analysis of retinal thickness in the ST/IT retina per C, M, and F subregions in patients with AD (n = 11), MCI (n = 6), and NC (n = 8–9). Two-way ANOVA: P_i–interactions, P_d–diagnostic groups, and P_r–retinal subregions. c, d Quantitative analysis of (c) brain atrophy severity scores in a subset of human donors with AD (n = 16), MCI (n = 9), or NC (n = 6), and (d) ST/IT retinal atrophy scores for an overlapping subset of patients with AD (n = 11), MCI (n = 6), or NC (n = 8 or 9). e–g Scatterplots display Pearson’s correlations between retinal thickness and retinal (e) Aβ₄₂ burden, (f) AβOi, and (g) S100β⁺ or GFAP⁺ macrogliosis. Color-filled dots represent these correlations in retinal mid-peripheral subregions. h Pearson’s correlation between retinal thickness and brain Aβ plaques. i Retinal thickness in subjects stratified based on high(H) or low(L) brain ATN-histopathology; extrapolated dotted-gray line marks the retinal thickness level for separating ATN^H from ATN^L individuals. j Pearson’s correlation between ST/IT retinal atrophy and MMSE cognitive score. k Representative retinal cross-sections labeled for cleaved caspase-3 (CCasp3; red) early apoptotic marker, GFAP (green), and DAPI nuclei (blue) in NC, MCI, or AD patients. Zoomed-in inserts are provided for representative images from the 2 AD patients to illustrate the presence of CCasp3⁺ cells within the INL and GCL. l Quantitative CCasp3-immunoreactive area in the ST/IT retina (n = 17). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by 1-way or 2-way ANOVA and Tukey’s post-hoc multiple comparison test

Fig. 6

Proteomic landscape of the retina and brain in AD. a Proteomics profiling of retinal (n = 6 AD; n = 6 NC) and brain (n = 10 AD; n = 8 NC) tissues. Heatmaps display detectable protein hierarchies from the ST/IT retina (temporal hemiretina), medial temporal gyrus (T.Cortex), hippocampus (Hipp.), and cerebellum; upregulated proteins are shown in pink and downregulated proteins in green. b Venn diagram depicting the number of overlapping differentially expressed proteins (DEPs) according to statistical significance (P < 0.05) and 1.2-fold change (FC) threshold criteria in the 4 analyzed CNS tissues; the number of common DEPs between paired CNS tissues (bold). c Volcano plots and top 20 up- or downregulated DEPs organized by FC (lowest P values highlighted in bold) in retinas and temporal cortices of AD vs. NC (DEPs marked by red circles). d DAVID-biological classification analysis displays top upregulated DEPs (pink) and top downregulated DEPs (green) in AD vs. NC retinas; lower blue bars represent magnitude of P values. Percentages indicate the fraction of each category of total up- or downregulated DEPs. e Pie chart of PANTHER-functional cluster analysis showing fraction and percentage of significant DEPs grouped by protein class category in retinas of AD patients vs. NC controls. f Ingenuity pathway analysis (IPA) of top up- and downregulated biological functions in AD vs. NC retinas. g Pearson’s (r) correlations between inflammatory/apoptotic-related DEPs in AD retinas identified by mass spectrometry and retinal Aβ_1–42 measured by ELISA in the same individuals