Overproduction of 42 Amino Acids Long Amyloid Beta Leads to Activation of Secretory Autophagy and Development of Drusen‐Like Structures Originating From Retinal Pigment Epithelium

ABSTRACT

Age‐related macular degeneration (AMD) is a global vision threatening disease affecting the macular region of the retina. AMD is classified into two forms: dry and wet AMD. There are no effective treatment options available for dry AMD (80% of cases). The cellular pathology includes oxidative stress and dysfunctional autophagy challenging the homeostasis of the retinal pigment epithelial (RPE) cells. Clinical findings include the formation of drusen deposits beneath the RPE cells consisting of 42 amino acids long amyloid beta (Aβ) among other components. However, the origin of drusen remains elusive. The 5xFAD (familiar Alzheimer's disease) mouse model of Alzheimer's disease produces abundant levels of Aβ making it an interesting model to study the possible relationship of Aβ to the formation of extracellular deposits and AMD‐like pathology. An immunohistology analysis of the 5xFAD mouse model showed accumulation of autophagic markers SQSTM1 (sequestosome 1) and ubiquitin in the RPE. Concurrently, the markers of secretory autophagy enabling the delivery of the intracellular material to the extracellular lumen were upregulated. Aβ, SQSTM1, ubiquitin, catalase, and TRIM16 (tripartite motif containing 16) shifted age‐dependently from intracellular origin to drusen‐like deposits beneath the RPE cells. Additionally, classical proteins secreted via secretory autophagy, IL‐1β (interleukin 1β), HMGB1 (high mobility group box‐1), and ferritin showed similar accumulation which became visible in fundus age‐dependently. These findings suggest a role for Aβ in the cellular pathogenesis of AMD. Furthermore, this model showed activated secretory autophagy pathway suggesting a role for Aβ in drusen‐like deposition formation.

The role of 42 amino acids long amyloid beta (Aβ) in the retinal pigment epithelium (RPE) was investigated using 5xFAD mice model overproducing Aβ. The RPE cells of 5xFAD mice showed increased oxidative stress, accumulation of autophagy‐destined proteins, increased expression of secretory autophagy markers, and increase in epithelial‐to‐mesenchymal transition ‐related markers. Clinical findings included accumulation of drusen‐like deposits and decreased vision. These findings make 5xFAD mice potential model for early AMD due to similar findings seen in AMD patient population.

1. Introduction

Age‐related macular degeneration (AMD) is the most common cause of blindness in Western countries, and the number of patients severely disabled by the disease is expected to increase more than 50% during the next 20 years [1]. It is a progressive and chronic disease of the central retina, stressing the lives of patients and ophthalmology clinics. AMD can be classified as dry (85%–90% of cases) and wet (10%–15%) forms [2, 3]. Both forms cause reduced central vision with loss of visual resolution essential for reading, driving, and recognizing faces, for example [4]. The current treatment for wet AMD is the intravitreal injections of anti‐vascular endothelial growth factor (VEGF) drugs, suppressing the growth and penetration of the vasculature from choroid to the retina and reducing retinal edema, clinical hallmarks of wet AMD [5]. Recently, the Food and Drug Administration (FDA) has approved complement inhibitors pegcetacoplan and avacincaptad pegol for the treatment of late‐stage dry AMD, geographic atrophy (GA) [6]. In contrast to the FDA, the European Medical Agency did not approve GA treatment due to limited effects on visual acuity. Therefore, dry AMD can still be considered an untreatable disease.

AMD can be considered as a multifactorial disease. Aging and genetics have been proposed to comprise most of the risk in the disease onset [7]. However, several lifestyle factors, such as smoking and high body mass index (BMI), have been associated with the disease prevalence [8]. At the cellular level, dysfunction of antioxidant defense system controlling oxidative stress and proteostasis network, including function of proteasomes and autophagy in the retinal pigment epithelial (RPE) cells at the center of the disease, have been linked to the AMD [9, 10, 11]. Insufficient/dysfunctional protein homeostasis coupled with oxidative stress leads to accumulation of degradation destined ubiquitinated proteins and autophagy receptor sequestosome 1 (SQSTM1/p62) tagged protein aggregates [12, 13]. Furthermore, accumulation of lipofuscin in the lysosomes due to photoreceptor outer segment (POS) phagocytosis hampers the function of degradation systems requiring lysosomal activity such as autophagy [14]. Also, a shift of epithelial cells to mesenchymal cells in a process called epithelial‐to‐mesenchymal transition (EMT) has been proposed to take place among the RPE cells during AMD progression [15, 16]. Increase of SNAIL, osteoblast(OB)‐cadherin, and vimentin have been observed in the RPE cells of oxidative stress‐related mouse model resembling AMD together with decrease in collagen 1 levels [17]. Furthermore, EMT‐related markers were elevated in the serum of wet AMD patients [18] and in the retina of postmortem human dry AMD samples [19] suggesting involvement of EMT to the disease pathology.

Despite the knowledge of the cellular events affecting disease onset and progression, the origin and reason for the accumulation of the extracellular protein deposits, drusen, between the RPE and Bruch's membrane limiting the retina from the choroid, remain elusive. However, drusen formation can be considered one of the key events leading to the disease phenotype, especially in the case of dry AMD [20]. Abnormal drusen size can predict AMD onset before affecting the sight of the patients [21, 22]. As the disease progresses, more severe drusen accumulation disturbs the function of the retina due to loss of contact to the choroid by the RPE cells maintaining the wellbeing of the entire retina. As a result, photoreceptor cells responsible for vision suffer from insufficient maintenance and start to degenerate, creating disturbance and loss of vision.

Drusen contains many similar components as seen in amyloidosis‐related diseases such as Alzheimer's disease (AD). For example, amyloid β (Aβ), inflammatory mediators, and protein degradation related factors are all found from the aggregating deposits in the AD and AMD [23, 24, 25]. Especially 42 amino acids long Aβ₄₂ (simply Aβ from now on) is known for its aggregating nature. The retina contains amyloid precursor protein (APP) which can be processed to aggregation prone Aβ via amyloidogenic pathway [26, 27]. The formation of Aβ requires activity of enzymes including beta‐site amyloid precursor protein cleaving enzyme 1 (BACE1, aka β‐secretase) and the γ‐secretase complex [26]. The genetic background of AD is strong [28], and the mutations observed on these genes in AD patients have led to the establishment of 5xFAD (familial Alzheimer's disease) mouse model overexpressing human APP with three FAD mutations and two FAD mutations in presenilin 1 (PSEN1), the catalytic unit of γ‐secretase, resulting in overexpression of Aβ with the major features of AD [29, 30]. The 5xFAD mouse model has been used widely in AD research. Despite the many similarities between AD and AMD, for example risk factors, strong genetic background, oxidative stress, and accumulation of extracellular deposits, there is no clear association between these diseases [31, 32, 33]. However, the overexpression of Aβ makes it an attractive model to study the formation of drusen‐like extracellular deposits and AMD‐related features [34].

Since dysfunction/insufficient function of proteasomal and autophagic degradation systems coupled with oxidative stress have been connected to the cellular pathology of AMD, there is a need for removal of accumulated protein waste and unneeded cellular organelles to sustain cellular homeostasis. Secretory autophagy (SA) offers an opportunity to relieve the intracellular burden of waste material and escape of certain intracellular proteins to the extracellular lumen non‐canonically [35, 36, 37]. In SA, the cargo contained cargo is released to the extracellular lumen via fusion with the plasma membrane instead of a proteolytic lysosome [37]. However, SA is a relatively unknown cellular pathway. Known molecular players of SA include TRIM16 (tripartite motif containing 16), SEC22B, syntaxin (STX) proteins, and Galectin‐8 guiding the SA destined cargo to the autophagosome and translocation of the autophagosome towards the plasma membrane for recognition by STX3/4 and SNAP23/29 (synaptosomal‐associated protein 23/29) and secretion [38, 39]. Interestingly, Aβ is known to be secreted via SA [40]. Also, SA seems to secrete and operate with several markers shared with EMT such as vimentin, interleukin 1‐beta (IL‐1β), ferritin, and high mobility group box‐1 (HMGB1), suggesting a relation between these two pathways [39, 40, 41].

This study sets out to characterize the 5xFAD mouse for AMD‐related molecular changes and phenotype. The expression of Aβ in the retina was examined with immunohistology together with markers related to protein aggregation, SA, and EMT. AMD‐resembling phenotype was observed with fundus photography and ocular coherence tomography (OCT) measurements coupled with electroretinogram (ERG) recordings to explore changes in vision.

2. Materials and Methods

2.1. Animals

In this study, female transgenic 5xFAD (n = 5–10) mice were used. 5xFAD mice express five human familial AD linked mutations in amyloid precursor (APP) and presenilin‐1 (PSEN1) genes: the Swedish (K670N/M671L), Florida (I716V), and London (V717I) mutations in APP, and the M146L and L286V mutations in PSEN1 (Alzforum [42]), driven by the mouse Thy1 promoter [43]. Wild‐type littermates were on the JAXC57BL/6 J background. The mice were housed in controlled temperature and humidity. The animals were maintained in a 12:12 h light–dark cycle and had food and water ad libitum.

Experiments were carried out in accordance with EU Directive 2010/63/EU for animal experiments and approved by the national Project Authorization Board and the studies were performed under project license (ESAVI/27769/2020) in compliance with the 3Rs principle. Association for Research in Vision and Ophthalmology (ARVO) statement for the Use of Animals in Ophthalmic and Vision Research was applied.

2.2. The Genotyping of 5xFAD Mice

The genomic mouse DNA was extracted from small ear samples by heating the tissues for 1 h at 95°C in 600 μL of 50 mM NaOH. The solutions were neutralized by adding 50 μL of 1 M Tris pH 8.0, and centrifuged for 6 min at 13 000 g.

The genotypes of 5xFAD mice were detected with a 4‐primer PCR. The presenilin 1 primer sequences were: 5′‐CTA GGC CAC AGA ATT GAA AGA TCT‐3′; 5′‐GTA GGT GGA AAT TCT AGC ATC ATC C‐3′. The genomic mouse DNA was studied with the following two primers: 5′‐AAT AGA GAA CGG CAG GAG CA‐3′; 5′‐GCC ATG AGG GCA CTA ATC AT‐3′.

Reactions included 2 μL of DNA extracted from mouse ears, 10 mM of each dNTPs, 1.5 mM of MgCl₂, 0.26 uL of presenilin 1 primers and 0.2 uL of genomic DNA, and 500 U of DreamTaq DNA polymerase (Thermo Fisher Scientific Waltham, MA, USA), in a volume of 50 μL. Samples were denatured at 94°C for 30 s, followed by 29 cycles at 94°C for 30 s, 54°C for 1 min, 68°C for 1 min, and a final extension at 68°C for 5 min, after that samples were in +4°C. All PCR products were separated using 1% agarose (Seakem LE, Cambrex) gel electrophoresis in 0.5× TBE buffer, visualized with Midori green Advance (Nippon genetics Europe), and utilized by UV light. The presence of rd8 (retinal degeneration 8) mutation was monitored by genotyping (Figure S1). All strains were negative for rd8 mutation.

2.3. ERG Recording

The function of retinal light sensitive cells was tested in scotopic and photopic conditions at 6‐, 9‐, and 12‐month‐old mice by ERG (Espion E3 System, Diagnosys UK Ltd., Cambridge, UK). The animals were dark‐adapted overnight. The mice were anesthetized using intraperitoneal injections with a mixture of 1.0 mg/kg body weight of medetomidine (Domitor vet 1 mg/mL, Orion Pharma, Espoo, Finland) and 75 mg/kg body weight of ketamine (Ketalar/Ketaminol 50 mg/mL, Pfizer Oy Animal Health, Espoo, Finland) and then placed on a thermal pad in the full‐field ERG dome, maintaining body temperature at approximately 38°C. The pupils were dilated with a mixture of tropicamide (Oftan Tropicamid 5 mg/mL, Santen Pharmaceutical Co. Ltd., Tampere, Finland) and phenylephrine hydrochloride (Oftan Metaoksedrin 100 mg/mL, Santen Pharmaceutical Co. Ltd., Tampere, Finland), and the surface of the cornea was moisturized with carbomer artificial tears (Viscotears 2.0 mg/g, Alcon a Novartis company). The recordings were conducted first in scotopic, followed by photopic protocol.

The ERG signals were recorded using a gold wire electrode placed on both eyes corneally. The signals were amplified with a band‐pass setting of 0.312–500 Hz and a sampling frequency of 2 kHz, except for scotopic ERG where the band‐pass was 1–300 Hz. A platinum needle reference electrode was placed subcutaneously on the forehead and a platinum needle ground electrode subcutaneously just superiorly from the tailbone. Both eyes were stimulated equally with ColorDome Ganzfeld full‐field ERG. The scotopic protocol included recording dark‐adapted responses to seven distinct stimulus intensities of blue light: 0.001, 0.003, 0.01, 0.03, 0.1, 0.3 and 1 Cd × s/m². Fifteen sweeps of 250 ms at each intensity were recorded with a delay of 10 s between each sweep. A 60 s light‐adaptation period in 20 Cd/m² white light (6500 K) was applied before the photopic responses. The animals were exposed to bright white background light for 5 min prior to the photopic measurements. The photopic protocol consisted of light‐adapted responses to four distinct stimulus intensities using white light: 1, 3, 10 and 20 Cd × s/m² in the presence of continuous background illumination of 20 Cd/m². Between the changes from one stimulus intensity to the subsequent one, an additional 20, 30 and 40 s of wait time was included. Twenty‐five sweeps of 300 ms were recorded with a delay of 5 s between each sweep. The baseline was set identically for all the recording protocols using the average voltage reading from the duration of 20 ms preceding the stimulus onset. The responses were then exported for off‐site analysis.

2.4. ERG Signal Processing and Feature Extraction

Post‐processing on the exported ERG data was performed with MATLAB (MathWorks MATLAB R2018b). For a‐wave analysis, the sweeps were averaged, and the a‐wave trough was determined as the lowest point of the signal following stimulus onset and preceding the rising phase of the b‐wave. The individual sweeps were then lowpass filtered using a 5th order Bessel filter with a stopband edge frequency of 60 Hz and then averaged for each stimulus intensity. The b‐waves were mapped on the averaged Bessel‐filtered signal by fitting a 2nd order polynomial in the surrounding of the highest value of voltage following the a‐wave. The width of the polynomial was dynamically set as twice the time elapsed from the a‐wave trough to the highest voltage point of the signal (initial estimate of the b‐wave peak). The polynomial fit was iterated five times or until no change in fit parameters occurred, always in the surrounding of the peak of the parabola. All peak fits were visually checked and adjusted manually if determined inaccurate. The b‐wave amplitude values were reported as the difference between the determined b‐wave peak and a‐wave trough. Similarly, the peak time of each wave was the time elapsed from the stimulus onset until the determined trough (a‐wave) or peak (b‐wave) of the signal.

2.5. OCT and Fundus Imaging

Following the ERG recordings, the retinal layers were determined with optical coherence tomography (OCT; Phoenix MICRON Image‐Guided OCT, Phoenix Technology Group LLC, Pleasanton, CA, USA) and retinal fundus images (Phoenix MICRON, Berkeley, CA, USA). In OCT, 10 repetitive images were acquired and averaged. The retinal layers from the averaged OCT images were evaluated by analysis with InSight image segmentation software (Voxeleron LLC, San Francisco, CA, USA). The acquired layer thickness data were then exported for off‐site statistical analysis. Total retinal thickness was inferred as the distance from the retinal nerve fiber layer to the retinal pigment epithelium (RPE) layer.

2.6. Eyes for Immunohistochemical Analysis

The mice were sacrificed with an overdose of a mixture of 1.0 mg/kg body weight of medetomidine (Domitor vet 1 mg/mL, Orion Pharma, Espoo, Finland) and 75 mg/kg body weight of ketamine (Ketalar/Ketaminol 50 mg/mL, Pfizer Oy Animal Health, Espoo, Finland) and cervical dislocation. The eyes were collected immediately and followed by fixation in 4% paraformaldehyde in 0.4 M phosphate buffer for overnight and continued with ethanol dehydration and embedded in paraffin. 5 μm thick parasagittal serial sections were cut with a microtome (SM2000 R, Leica, Heidelberg, Germany) for immunohistochemical analysis. The cut tissue sections were deparaffinized by using xylene and rehydrated. The antibodies used are in Table 1.

TABLE 1.

Antibodies used in immunohistochemical analysis.

Protein	Manufacturer	Cat#	Host	Dilution	Pretreatment
4‐HNE	LS‐Bio	LS‐C68182	Rabbit pAb	1/200	—
Beta‐amyloid	Sigma	MABN10	Mouse	1/4000	CB
Collagen‐1	SantaCruz	sc‐393 573	Mouse mAb	1/100	CB
Catalase	Abcam	ab16731	Rabbit	1/100	—
Ferritin	Thermo	MA5‐32244	Rabbit mAb	1/300	—
Galectin 8	Thermo	PA5‐50965	Rabbit	1/100	CB
HMGB1	Abcam	ab79823	Rabbit mAb	1/200	CB
IL‐1b	R&D systems	AF‐401‐S	Goat	1/100	—
SQSTM1/p62	Abcam	ab76340	Rabbit pAb	1/100	CB
SEC22B	Abcam	ab181076	Rabbit	1/200	—
SNAIL	SantaCruz	sc‐271 977	Mouse mAb	1/100	CB
TRIM16	Thermo	PA5‐110515	Rabbit	1/100	CB
Ubiquitin	DAKO	Z0458	Rabbit pAb	1/200	—
Vimentin	SantaCruz	sc‐373 717	Mouse mAb	1/100	—
RPE‐65	Abcam	ab78036	Mouse	1/250	—

Abbreviation: CB, citrate buffer treatment.

Sections for Snail, beta‐amyloid, SQSTM1, ubiquitin and collagen 1 were pre‐treated with 1× citrate buffer for 5 min in 90°C. All the sections were further processed according to the same protocol. Sections were permeabilized twice for 5 min with 0.1% Triton‐TBS. After rinsing, sections were blocked with 20% goat serum in TBS for 30 min. Before primary antibody staining, the removal of autofluorescence was performed with TrueBlack (Biotium) according to the manufacturer's instructions. Primary antibodies were incubated overnight at 4°C with 1% goat serum in TBS. After primary antibody incubation, the sections were first rinsed and then incubated in dark with secondary antibody Alexa Fluor (594) diluted 1:500 in TBS for 3 h. After washing the sections three times in dark, fluorescent nuclear marker DAPI (4,6‐diamidino‐2‐fenilindole, dihydrochloride) (Sigma Aldrich, St. Louis, MO, USA) was diluted 1:10000 in TBS and incubated for 30 min in RT. After rinsing, slides were covered with Mowiol mounting media (Sigma Aldrich, St. Louis, MO, USA) and stored in the dark at room temperature. Double staining with RPE65 (retinal pigment epithelium 65 kDa) and TRIM16 as well as TRIM16 and Aβ (W02) was performed by mixing the primary and secondary antibodies in concentrations as stated in the Table 1 and above, respectively. Antibody selection was based on manufacturers' specificity and validation tests for use in immunohistochemistry and existing scientific articles.

2.7. Hematoxylin–Eosin (H&E) Staining of Paraffin Sections

Sections were deparaffinized by immersion in xylene. After that, sections were rehydrated through graded ethanol; slides were carried from absolute ethanol to 94% ethanol. After rehydration, the slides were rinsed in distilled water. Staining was performed by immersing the sections in Harris hematoxylin for 5 min, which stains cell nuclei. The slides were then rinsed under running tap water to remove hematoxylin. After rinsing, there were a quick four dips in 1% hydrochloric acid in ethanol, followed by another rinse in running tap water. Counterstaining was achieved by immersing the slides in 1% eosin in alcohol for 30 s and after that the sections were dehydrated by passing them through 94% ethanol and absolute ethanol. Slides were covered with DPX New mounting media and stored at room temperature.

2.8. Fluorescent and Light Microscopy and Imaging

Eye sections were examined with a fluorescent microscope (Zeiss Axio Imager M2, Zeiss, Göttingen, Germany). Images were recorded by using a 63× oil immersion objective. The microscopy settings were identical for all scans and kept constant during imaging. Representative areas per sample were selected for collecting RPE cells from each sample close to the optic nerve. Ten cells were randomly selected from each group (WT and 5xFAD, n = 6–8) for counting. Under a microscope, the Bruch's membrane appears as a faint streak under the RPE cells. The signal that was localized clearly to the choroid was left outside from the analysis. The signal around the RPE cells was calculated into the RPE cell layer signal; the signal on the Bruch's membrane was calculated into the basal lamina/Bruch's membrane layer signal. The images were used to measure mean gray values by using ImageJ. Bright field images from H&E stained sections were captured using Zeiss Axio Imager M2 with 40× magnification. The images were visually inspected for retinal layers as well as the Bruch's membrane‐choroid axis to evaluate possible differences.

2.9. Transmission Electron Microscopy

In transmission electron microscopy (TEM), the mice eyes were pre‐fixed with 4% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4) with 1% glutaraldehyde for 5–7 days at +4°C. After 2 × 15 min washing in 0.1 M cacodylate buffer, the samples were post‐fixed in 1% osmium tetroxide in 0.1 M cacodylate buffer (pH 7.4) for 2 h and again washed with cacodylate buffer for 2 × 10 min and 5 min in distilled water prior to standard ethanol dehydration. Subsequently, the samples were infiltrated and embedded in LX‐112 resin (Ladd Research, Williston, VT, USA). Polymerization was carried out at 60°C for 48–72 h.

One‐micron semi‐thin sections were cut with a Leica EM UC7 microtome, stained with 1% toluidine blue, and examined with a light microscope to find the localization of interest for further TEM sectioning. Next, the 70 nm thick sections (cut with Leica EM UC7, Wetzlar, Germany) were stained with 1% uranyl acetate for 30 min and with lead citrate for 2 min. The sections were examined with a JEM‐2100F TEM (Jeol, Tokyo, Japan) at 200 kV. Six cells were randomly selected from each mouse (WT and 5xFAD mice aged 6 and 12 months, n = 36) for counting. Autophagic structures were manually counted by using ImageJ.

2.10. Data Analysis

Statistical analyses were performed with IBM SPSS Statistics for Windows (version 26.0, Armonk, NY, USA). Shapiro–Wilk test of normality was conducted and due to data not being in a normal distribution (p < 0.05), non‐parametric Mann–Whitney U‐test was used for comparisons between groups. Data are presented as the mean ± standard deviation (SD).

The statistical analysis of the ERG results was done with R (version 3.5.3). The regular two‐way analysis of variance (ANOVA) was used to infer statistical significance for each stimulus intensity with Bonferroni correction. The main effect between each treatment group and the genotype was determined by two‐way ANOVA using the genotype and stimulus intensity as independent variables. If significance was determined, the test was followed by Bonferroni post hoc test with multiple pairwise comparisons using Student's t‐test. The OCT results were analyzed using Student's t‐test using R.

3. Results

3.1. 5xFAD Mice Display Accumulation of 42 Amino Acids Long Aβ to the Proximity of Bruch's Membrane in a Time‐Dependent Manner

The accumulations of amyloid peptides in the RPE cells were analyzed from 3‐, 6‐, and 12‐month‐old 5xFAD and WT animals to investigate their composition, time‐dependent quantity and localization (Figure 1A). W02‐antibody recognizes all APP fragments containing amino acids 4–10 of the Aβ region of the human APP, including full length APP, amyloid precursor protein α (sAPPα), and C‐terminal fragment β (CTFβ), whereas Aβ₄₂‐antibody recognizes specifically C‐terminal fragments of Aβ₄₂. It is possible to specify the molecules by double staining with APP N‐terminal 22C11 (MAB348, Merck) and APP –C‐terminal (A8717, Sigma) antibodies. If the signal of 22C11 is positive but the signal of APP‐CTF is negative, the stained molecules are sAPPα or sAPPβ. If there is no signal of 22C11 and APP‐CTF is positive, the identity of the fragment depends on the W02 signal. If W02 is positive, the stained protein is CTFβ, but CTFα if W02 is negative. If both APP‐CTF α and β are positive, the signal is full‐length APP. Based on the immunostaining with these three antibodies, we could identify W02 positivity in the WT animals as full‐length APP‐protein (positive signal for both 22C11 and APP‐CTF), whereas the W02 signal in the 5xFAD mice originated exclusively from Aβ₄₂ (negative signal for 22C11 and APP‐CTF, positive for Aβ42‐antibody) (Figure S2). The quantitative analysis was performed using only the W02‐antibody.

FIGURE 1.

Levels of APP/Aβ, ubiquitin, and SQSTM1 are elevated in the RPE cells and shifts towards basal localization and beneath the Bruch's membrane in a time‐dependent manner. (A) The APP level in WT was highest in the RPE cells at the age of 3 months. The APP levels are localized to the basal lamina/Bruch's membrane towards 12 months of age. Similar shift of Aβ was observed in the 5xFAD mice. The signal of Aβ in the 5xFAD mice was localized to structures beneath the Bruch's membrane at the age of 12 months (white circle). (B) The levels of ubiquitin showed elevated levels in the RPE cells in both WT and 5xFAD mice with significantly higher intensity observed in the 5xFAD mice at the age of 3 months. The ubiquitin signal was absent in the WT mice and significantly elevated in the 5xFAD mice at the age of 6 months. At the age of 12 months, WT and 5xFAD mice showed similar levels of ubiquitin in the basal lamina/Bruch's membrane with clear accumulation of ubiquitin to the structures located beneath the Bruch's membrane (white circles). (C) Autophagy cargo receptor SQSTM1 showed elevated levels in both WT and 5xFAD in the RPE cells at 6 months old. The levels of SQSTM1 shifted from RPE cells to the basal lamina/Bruch's membrane at the age of 12‐months. WT mice showed higher accumulation of SQSTM1 to the basal lamina/Bruch's membrane at the age of 12 months whereas the 5xFAD mice showed high accumulation of SQSTM1 to large structures (white circle) beneath the Bruch's membrane. Bruch's membrane (white arrow), RPE cells (yellow arrow). Scale bar 10 μm, *p < 0.05, **p < 0.01, ****p < 0.001, Mann–Whitney U‐test.

The quantitative analysis was performed from two locations: the RPE cell excluding basal lamina/Bruch's membrane and RPE cell basal lamina/Bruch's membrane. Three‐month‐old 5xFAD mice showed similar levels of Aβ in the RPE cells as APP in WT mice (Figure 1A). Both mouse strains lacked basal lamina/Bruch's membrane localized amyloid peptides at 3 months of age. In 6‐month‐old mice, amyloid peptides were found inside the RPE cells and as basal lamina/Bruch's membrane; the amyloid levels were higher in 5xFAD mice at both locations. In 12‐month‐old mice, the amyloid levels disappeared from inside the RPE cells of both mouse strains but remained localized at the basal lamina/Bruch's membrane. However, the amyloid levels of both mouse strains were similar. Interestingly, 12‐month‐old 5xFAD mice showed deposit structures with intensive Aβ signal beneath the Bruch's membrane. As a summary, WT and 5xFAD mice seemed to express unprocessed APP and Aβ, respectively, as age‐dependent localization. Both forms of amyloid peptides were shifted from inside the RPE cells to the basal surface of RPE cells/Bruch's membrane during aging; Aβ in the 5xFAD mice accumulated beneath the Bruch's membrane.

3.2. Protein Ubiquitination and Aggregation Markers Shift From RPE Cells to Deposit Structures Beneath the Bruch's Membrane in a Time‐Dependent Manner

Since protein aggregation and autophagy dysfunction are well described in cellular pathology of AMD, the levels of protein ubiquitination and SQSTM1 were analyzed from 5xFAD and WT mice. The level of protein ubiquitination inside the RPE cells was significantly higher in 3‐month‐ (p < 0.02) and 6‐month‐old (p < 0.,014) 5xFAD mice (Figure 1B). The area of ubiquitin signal was significantly higher in 5xFAD suggesting higher aggregation of ubiquitinated proteins (data not shown). Basal lamina/Bruch's membrane localized ubiquitin signal appeared at 6‐month‐old (p < 0.0001) 5xFAD retina, whereas the WT mice showed only intracellular ubiquitin levels. Also, large apically localized ubiquitin positive structures were seen in 6‐month‐old 5xFAD mice. In 12‐month‐old retinas, both mouse strains expressed basal lamina/Bruch's membrane localized ubiquitin signal without significant difference in the quantity or area of the signal. Both strains lacked intracellular signal for ubiquitin. However, intensive ubiquitin signal was localized beneath the Bruch's membrane of 5xFAD mice in a similar manner as Aβ previously. Protein aggregation marker SQSTM1 was analyzed from 6‐ and 12‐month‐old mice retinas (Figure 1C). In 6‐month‐old mice, the levels of SQSTM1 did not show significant differences between mouse strains. However, the shift of SQSTM1 localization from RPE cells to basal lamina/Bruch's membrane was seen from 6‐ to 12‐month‐old mice. A similar shift was already seen with Aβ and ubiquitin. Interestingly, 12‐month‐old 5xFAD mice showed significantly lower levels of SQSTM1 (p < 0.0177) signal localized to basal lamina/Bruch's membrane. However, most SQSTM1 signal was seen beneath the Bruch's membrane in 5xFAD mouse retinas. Autophagy reporter mouse model CAG‐RFP‐EGFP‐LC3 was combined to 5xFAD mice strain to visualize autophagic structures at the age of 6 months. WT mice showed only a few autophagosomes whereas CAG‐RFP‐EGFP‐LC3/5xFAD mice RPE cells contained autolysosomes (Figure S3) indicating enhanced autophagy flux. Furthermore, oxidative stress has been proposed as one of the key elements of cellular pathology of AMD. Interestingly, the 5xFAD mice showed less oxidative damage measured as 4‐HNE levels (p < 0.0001) compared to WT animals at the age of 6 months (Figure S4A). Similarly, antioxidant enzyme catalase showed decreased expression levels (p < 0.023) at the same age in both mouse strains (Figure S4B). Six months later, 4‐HNE and catalase levels were significantly higher (p < 0.0001) in 5xFAD mice suggesting higher oxidative stress. The level of catalase seemed to decrease over time in WT mice RPE cells whereas accumulation of catalase beneath the Bruch's membrane to a deposit structure was observed in 5xFAD mice retinas. In summary, WT and 5xFAD mice showed a shift of APP, Aβ, ubiquitinated proteins, SQSTM1 and catalase localization from intracellular to basal lamina/Bruch's membrane. However, only the 5xFAD mice showed an accumulation of these markers beneath the RPE cells in drusen‐like deposits.

3.3. EMT‐Related Markers Are Elevated in 5xFAD Mice RPE Cells

EMT has been connected to AMD and its activation detected in rodent models of AMD. Therefore EMT‐related markers vimentin, collagen 1, SNAIL, and ferritin were measured from 6‐month‐old mice retinas (Figure 2). The 5xFAD mice RPE cells showed significantly increased levels of intracellular collagen 1 (p < 0.0001) and SNAIL (p < 0.0001), whereas ferritin (p < 0.0001), SNAIL (p < 0.0001), collagen 1 (p < 0.0001), and vimentin (p < 0.0002) levels were significantly elevated at the basal lamina/Bruch's membrane. The signal of SNAIL was absent in WT mice, whereas ferritin was detected only from the basal lamina/Bruch's membrane of 5xFAD mice. At the same time, a shift of APP, Aβ, catalase, ubiquitinated proteins, and SQSTM1 from intracellular origin to the proximity of Bruch's membrane were observed, suggesting a relationship on dynamics of these proteins. Interestingly, vimentin and ferritin were found to localize to the proximity of the Bruch's membrane in 5xFAD mice retinas; both are suggested to be secreted via SA.

FIGURE 2.

Epithelial‐to‐mesenchymal transition markers are elevated in the RPE cells of 5xFAD mice. (A) The level of collagen 1 was significantly higher in the RPE cell and the basal lamina/Bruch's membrane locations in the 5xFAD mice at the age of 6 months. (B) The level of vimentin was significantly higher in the basal lamina/Bruch's membrane in the RPE cells of 5xFAD mice at the age of 6 months. (C) The level of SNAIL was absent in the WT mice RPE cells and basal lamina/Bruch's membrane. The 5xFAD mice showed increase in SNAIL intensity in the RPE cells and basal lamina/Bruch's membrane locations at the age of 6 months. (D) The level of ferritin was absent in the WT mice RPE cells and basal lamina/Bruch's membrane. The 5xFAD mice showed significant increase of vimentin signal in the basal lamina/Bruch's membrane and beneath the Bruch's membrane (white circle) at the age of 6 months. The signal of ferritin was absent in the RPE cells of 5xFAD mice. Bruch's membrane (white arrow), RPE cells (yellow arrow). Scale bar 10 μm, ***p < 0.005, ****p < 0.001, Mann–Whitney U‐test.

3.4. HMGB1 Levels Were Increased at the Bruch's Membrane and IL‐1β Localize to Deposit Structures Suggesting Activated Secretory Autophagy (SA)

Since EMT analysis showed possible secretion of vimentin and ferritin from the RPE cells, and the Aβ, ubiquitin, and SQSTM1 were found from deposits beneath the Bruch's membrane, we analyzed HMGB1 and IL‐1β levels due to their established secretion via SA. The retinas of 12‐month‐old 5xFAD mice showed significant upregulation of HMGB1 (p < 0.0001) and IL‐1β (p < 0.0001) in proximity to Bruch's membrane suggesting their secretion via SA (Figure 3A,B). Furthermore, both markers seemed to accumulate beneath the RPE cells similarly to Aβ, ubiquitin, SQSTM1, catalase, and ferritin as shown earlier.

FIGURE 3.

The levels of secretory autophagy markers are elevated in the RPE of 5xFAD mice. (A) The signal of HMGB1 was present in the basal lamina/Bruch's membrane of 5xFAD mice at the age of 12 months. The signal localized to the structures beneath the Bruch's membrane (white circle). The HMGB1 signal was absent in the WT mice and in the RPE cells of 5xFAD mice. (B) The signal of IL‐1β showed higher levels in the RPE cells of 5xFAD mice in comparison to the WT mice at the age of 12 months. The signal of IL‐1β was present in the structures localized beneath the Bruch's membrane (white circle). (C) The secretory autophagy receptor TRIM16 levels were absent in the WT mice whereas a clear increase was observed in the RPE cells of 5xFAD mice at the age of 12 months. The TRIM16 signal was present in structures beneath the Bruch's membrane (white circle). (D) The TRIM16 cooperating partner Galectin‐8 levels were significantly higher in the RPE cells of 5xFAD mice at the age of six and 12 months. (E) The cargo delivery‐related protein SEC22b levels were significantly higher in the RPE cells of 5xFAD mice at the age of 6 months. The intensity levels between WT and 5xFAD mice did not show difference at the age of 12 months. The number of positive punctae in the RPE cells of 5xFAD mice was clearly higher in the 5xFAD mice at the age of six and 12 months (not analyzed). Bruch's membrane (white arrow), RPE cells (yellow arrow). Scale bar 10 μm, **p < 0.01, ****p < 0.001, Mann–Whitney U‐test.

3.5. Secretory Autophagy Mediators Are Upregulated in the 5xFAD Retina

Since classical cargo of SA, HMGB1, ferritin, vimentin, IL‐1β, and Aβ was found from basal lamina/Bruch's membrane localized and in the deposits beneath the Bruch's membrane, SA mediators were analyzed from the RPE cells of both mouse strains. Cargo receptor TRIM16 levels were observed only in the 5xFAD mice RPE cells (6 months p < 0.0001, 12 months p < 0.0001) and in deposits beneath the Bruch's membrane (Figure 3C). The interacting partner of TRIM16, galectin‐8, was present at higher levels in 5xFAD mice than in WT mice RPE cells (6 months p = 0.0055, 12 months p = 0.0042) (Figure 3D). The localization of galectin‐8 was restricted intracellularly. The levels of Sec22b, connecting cargo receptor TRIM16 and cell membrane localized SNAP23/29 delivering the cargo to the extracellular space, were also increased in the 5xFAD mice RPE cells at 6 months of age (p < 0.0001) (Figure 3E). There was no significant difference between the signal intensities in WT and 5xFAD mice at the age of 12 months. However, the number of positive punctae in 5xFAD RPE cells was clearly higher at the age of 6 and 12 months. The localization of Sec22b was more basal and intracellular.

3.6. Double Staining With RPE65 + TRIM16 and TRIM16 + Aβ Showed the Shift of TRIM16 Outside From the RPE Cells and Co‐Localization of TRIM16 and Aβ Inside the RPE Cells and Drusen‐Like Deposits

To verify the shift of SA cargo receptor TRIM16 from inside the RPE cell to the Bruch's membrane and drusen‐like deposits, the RPE cells were double stained with RPE65 and TRIM16 (Figure 4A). The RPE65 stains only the RPE cell layer aiding the locational visualization of TRIM16 signal. The WT mice did not show any signal for TRIM16, whereas 5xFAD mice showed signal in the basal side of the RPE cells and basal lamina/Bruch's membrane at 6 months old. The signal in the basal lamina/Bruch's membrane was stronger at the age of 12 months, and the co‐localization in the basal side of RPE was decreased. Interestingly, 12 months old 5xFAD mice showed clearly decreased RPE65 staining.

FIGURE 4.

Double staining with RPE65 + TRIM16 and TRIM16 + Aβ showed localization of TRIM16 outside the RPE cell and co‐localization of TRIM16 with Aβ. (A) The double staining with secretory autophagy receptor TRIM16 (green‐fluorescent dye) and RPE65 (red‐fluorescent dye) in WT and 5xFAD mice retinas in 6‐ and 12‐month timepoints. (B) The double staining with Aβ (red‐fluorescent dye) and TRIM16 (green‐fluorescent dye) showed in 6‐ and 12‐month‐old 5xFAD mice showed strong co‐localization especially in drusen‐like structures beneath the RPE cells. GCL = ganglion cell layer, INL = inner nuclear layer, IPL = inner plexiform layer, ONL = outer nuclear layer, OPL = outer plexiform layer, POS = photoreceptor outer segments, RPE = retinal pigment epithelium, BM = Bruch's membrane (white arrow), Ch = Choroid, purple arrow = co‐localization. Scale bar: 20 × 50 μm, 63 × 10 μm.

Since TRIM16 acts as a cargo receptor for SA, double staining with Aβ and their co‐localization would shed light to the possible interaction of these proteins. Some co‐localization can be seen at 6 months old 5xFAD mice RPE cells as well as basal lamina/Bruch's membrane (Figure 4B). At 12 months of age, 5xFAD mice showed strong co‐localization in drusen‐like structures beneath the RPE cells as well as inside the RPE cells.

3.7. Ultrastructural Analysis Revealed Larger Autophagosome in the 5xFAD Mice

The quantitative analysis of autophagic structures, autophagosomes and autolysosomes, did not show differences between the average number of structures (Figure 5). However, the area of the autophagosomes in the RPE cells increased significantly in the 5xFAD mice at the age of 12 months (p = 0.0052). This is expected as the lysosomal degradation system seems to be hampered/overwhelmed by oxidative stress, Aβ accumulation, and autophagic degradation destined material accumulation. Also, the size of the autolysosomes showed a trend towards larger structures in the RPE cells of 5xFAD mice, which can be related to the malfunctioning lysosomal degradation.

FIGURE 5.

Ultrastructural analysis of autophagic structures in WT and 5xFAD mice RPE cells at the age of six and 12 months. The average area of autophagosomes (pink circles) were higher in the RPE cells of 5xFAD mice at the age of 12 months. The average area of autolysosomes (orange circles) showed a trend towards larger structures in the 5xFAD mice RPE cells at both ages. Abnormal Bruch's membrane (yellow arrow) morphology was seen in 5xFAD mice at both ages. The number of autophagic structures did not show statistically significant difference. Mitochondria = blue arrowhead, melanosome = purple arrowhead, the scale bar indicates 1 μm. **p < 0.01, Mann–Whitney U‐test.

3.8. Fundus Imaging, Optical Coherence Tomography, and Hematoxylin & Eosin Staining of 5xFAD Mice Retina Revealed AMD Resembling Accumulation of Deposits, Alterations in Bruch's Membrane, and Downward Trend in Retinal Thickness

Since the accumulation of deposits containing key molecules found in human drusen was observed beneath the RPE cells of 5xFAD mice, fundus imaging and OCT were performed to visualize and confirm the occurrence of these deposits from 6‐, 9‐, and 12‐month‐old mice. At 6 months of age, WT mice did not show obvious signs of drusen‐like deposits in the fundus photographs, but later time points revealed accumulation of some small deposits (Figure 6A). In contrast, the 5xFAD mice fundus started to show drusen‐like deposits already at the age of 6 months with increasing volume and number towards later ages.

FIGURE 6.

Fundus photographs, H&E staining, and OCT scans of the 5xFAD mice retina showed age‐dependent accumulation of drusen‐like deposits and abnormal Bruch's membrane without changes in total retinal thickness. (A) Fundus photographs taken at 6, 9, and 12 months of age showed the appearance of drusen‐like deposits (yellow circles and arrows) in the 5xFAD mice. The WT mice developed some smaller deposits as a result of normal aging. (B) Representative (40×) microscope images of H&E‐stained retinas from 12‐month‐old WT and 5xFAD mice. GCL = ganglion cell layer, INL = inner nuclear layer, IPL = inner plexiform layer, ONL = outer nuclear layer, OPL = outer plexiform layer, POS = photoreceptor outer segments, RPE = retinal pigment epithelium, BM = Bruch's membrane (blue arrow), Ch = choroid, red circles = abnormal Bruch's membrane morphology. Scale bar 20 μm. (C) Representative OCT images of WT and 5xFAD mice retina (scale bar = 100 μm). Arrow = deposit like structure. (D) RPE and total retinal thickness were preserved in both groups and age ranges of 6 and 12 months, though 5xFAD mice showed a downward trend in total retinal thickness. SD = standard deviation, mo = months.

The H&E staining of retinal sections did not show remarkable changes in retinal layers between 5xFAD and WT mice (Figure 6B). However, the Bruch's membrane showed waviness and rupture which could be seen mostly in the retinae of 5xFAD mice. Similar waviness of the Bruch's membrane and merging of the RPE‐Bruch's membrane‐choroid layers was seen in the OCT scans where fundus images showed yellowish drusen‐like deposits.

The total retinal thickness and RPE thickness yielded no statistically significant changes between the phenotypes or the age groups (Figure 6C,D). Although it seems that the total retinal thickness in WT mice might be more preserved in aging than in the 5xFAD, which had a trend towards reduced total retinal thickness. In both groups, the RPE thickness seemed slightly reduced by the age of 12 months.

3.9. The Scotopic ERG Revealed That WT Mice Have More Preserved Visual Responses at Older Age Than 5xFAD Mice

With increasing age, the a‐wave amplitudes were significantly reduced from 3 to 12 months of age in both WT (p < 0.01) and 5xFAD (p < 0.001) mice (Figure 7A–D). Additionally, 5xFAD mice at 12 months of age had significantly lower a‐wave amplitudes than WT mice at the same age (p < 0.001). The higher amplitude of WT compared to 5xFAD was only pronounced enough at the age of 12 months to yield statistically significant difference (p < 0.01). Similar effects were observed in the b‐wave amplitudes. While the trend seemed to be that 12‐month aged mice had lower b‐wave amplitudes, the 6‐month age was interesting since the b‐wave amplitudes reduced to 6 months but then increased by 12 months. In 5xFAD mice, the changes were statistically significant in both comparisons of 3 months to 6 months (p < 0.01) and 3 months to 12 months (p < 0.001). Similarly, for the WT, the changes were significant between ages of 3 months and 6 months (p < 0.001) as well as 3 months and 12 months (p < 0.001). Additionally, the WT mice had higher amplitude than the 5xFAD in age groups of 3 (p < 0.001) and 12 months (p < 0.001).

FIGURE 7.

Scotopic and photopic ERG responses of 5xFAD and WT mice at ages of 3, 6 and 12 months. (A) The scotopic a‐wave amplitudes of 12‐month‐old 5xFAD mice were found to be significantly reduced in comparison to the WT mice. (B) The scotopic a‐wave peak time was marginally reduced in 12‐month‐old 5xFAD mice, and this was seen only in the dimmest flash response. (C) Across the dim flash responses, the scotopic b‐wave amplitude was significantly more impaired in the 5xFAD mice at the age of 3 months and 12 months. (D) In aged 5xFAD mice (6 months and 12 months), the scotopic b‐wave peak time was significantly reduced in selected higher intensity flash responses. (E) The photopic b‐wave amplitudes tended to be higher in especially aged 5xFAD mice (6 and 12 months) compared to WT, while the trend seemed to be opposite for younger (3 month) individuals. (F) In photopic b‐wave peak time, no statistical significance was found between the phenotypes and visually inspecting no clear trend could be seen *p < 0.05, **p < 0.01 and ***p < 0.001, Student's t‐test.

When observing the kinetics of the ERG response by assessing peak times of a‐ and b‐waves, we found significant changes in aging mice. The a‐wave peak times were significantly shortened in 12‐month‐old 5xFAD mice in comparison to their 3‐month‐old counterpart (p < 0.001). The b‐wave peak times in 5xFAD mice were significantly reduced by 6 months (p < 0.01) and even more by 12 months (p < 0.05), compared to the ages of 3 and 6 months, respectively. Interestingly, in the WT mice, the peak time first prolonged from 3 months to 6 months (p < 0.001) and then it shortened close to the baseline level by the age of 12 months (p < 0.001). In all age groups the differences between 5xFAD and WT mice were statistically significant (p < 0.05).

3.10. In the Photopic ERG, the Effects of Aging Were Seen in Both Phenotypes but no Difference Among the Phenotypes Was Detected

In photopic ERG recordings, 5xFAD mice experienced significantly reduced b‐wave amplitudes at the age of 12 months compared to 6 months (p < 0.01) (Figure 7E,F). However, this effect was only clearly visible at the higher flash intensities. Conversely, the amplitudes were higher in the 6‐month than in the 3‐month group (p < 0.05). Among WT mice, there was a statistically significant difference between the age groups of 3 and 12 months (p < 0.01) as well as between 6 and 12 months (p < 0.001), showing reduced amplitudes at the age of 12 months. The main effect between the phenotypes reached no statistical significance, though in some individual flash intensities differences could be visualized. For instance, in aged 5xFAD mice (6 and 12 months), there was a trend towards higher b‐wave amplitude values in comparison to the WT phenotype.

In comparing the b‐wave peak times, 5xFAD mice displayed significantly shorter b‐wave peak times at the age of 12 months compared to 3 months (p < 0.01). In WT mice, there was a statistically significant difference between the age groups of 3 and 6 months (p < 0.01) as well as between 3 and 12 months (p < 0.001), yielding shorter peak times in the 6‐ and 12‐month groups. The main effect between the phenotypes reached no statistical significance.

4. Discussion

The origin of drusen deposits in AMD patient retinas has remained unclear. A lot of emphasis has been put on a qualitative and quantitative analysis of drusen composition to reveal their origin. However, the qualitative composition of drusen originating from several different sources and the lack of potential secretory pathways delivering material to the extracellular lumen have complicated theories of their possible origin [44, 45]. Bergen et al. [44] classified 89 proteins from drusen and found ~11% to originate from RPE/photoreceptor cells (~9% uniquely from RPE cells), ~61% from choroid/blood, and the rest from unknown origin. In total, ~45% of the studied proteins were localized to cytoplasm, nucleus, or plasma membrane, the rest were extracellular matrix proteins. Eight of 10 uniquely RPE cell expressed proteins were either cytoplasm (7) or plasma membrane (1) localized proteins. Thus, many of the components of drusen seem to originate from an intracellular source rather than being secreted conventionally to the extracellular matrix. A decade ago, a new form of secretory pathway, SA, has been suggested to deliver leaderless intracellular proteins to the extracellular lumen [37, 46]. The specific cargo for the pathway includes IL‐1β, possibly the most well‐characterized protein secreted via SA, vimentin, ferritin, HMGB1 and, Aβ [37, 39]. Since Aβ can be found from drusen specimens from AMD donor eyes, is well known for its aggregating nature, and having a role in AMD development, studies on human Aβ overexpressing mice model with known accumulation of Aβ deposits in the retina becomes of interest.

Several Alzheimer's disease models have been studied previously in relation to AMD, including the 5xFAD mouse model [25, 47]. The previous findings with 5xFAD mice include abnormal Bruch's membrane and choroid, increased lipofuscin accumulation in the RPE cells, and the presence of basal linear and basal laminar deposits [34, 48, 49, 50]. However, previous studies have focused more on the neural retina excluding the RPE cells and retinal electrophysiology. In this study, the focus was kept on molecular and pathological changes observed in the RPE layer or beneath the RPE cells.

WT mice seemed to express only full‐length APP. The majority of APP was localized inside the RPE cells at the age of 3 months with increasing localization to the basal lamina/Bruch's membrane over time. At the age of 12 months, all APP signal was found from basal lamina/Bruch's membrane. Interestingly, 5xFAD mice seemed to express only Aβ with a similar localization pattern from inside the RPE cell to the extracellular space. A similar time‐dependent shift from intracellular to extracellular was found also for protein degradation‐related markers ubiquitin and SQSTM1. Furthermore, Aβ, ubiquitin, and SQSTM1 immunohistochemistry showed positive signals in large drusen‐like deposits found only from 5xFAD mice retinas at the age of 12 months. Furthermore, the autophagy reporter mouse model combined with 5xFAD mice showed enhanced autophagic degradation at the age of six months. This suggests still functional autophagy flux, which may be overwhelmed during aging. This is seen as a lack of SQSTM1 signal outside of the cell in immunohistochemical analysis and drusen‐like structures in the fundus photographs at the age of 6 months. Furthermore, TEM data showed a trend towards larger autolysosomes in the 5xFAD mice at the age of 6 months whereas the autophagosomes remained at the same size as in WT mice. At the age of 12 months, the size of the autophagosomes became statistically larger and the localization of the autophagy cargo receptor SQSTM1, usually degraded during the autophagy, localized outside of the cell. Interestingly, the antioxidative enzyme catalase showed a similar pattern. These deposits were located well beneath the RPE cells in the choroid possibly due to a breach of the Bruch's membrane and not between the Bruch's membrane and RPE as seen in AMD patient samples. However, the use of murine models often results in drusen‐like deposits that are not localized in a comparable manner as in the human retina [51]. The accumulation of extracellular debris to the basal laminar‐like deposits can be considered normal aging [52]. However, larger drusen‐like deposits found from 5xFAD mice suggest a disease phenotype in connection to the Aβ accumulation.

The presence of large drusen‐like deposits in the retinas of 5xFAD mice with immunohistochemical positivity to markers usually connected to the autophagic degradation process (Ubiquitin and SQSTM1) and Aβ known to be secreted via SA [38, 53, 54], closer inspection to the possible activation of SA was performed. Possibly the best characterized cargo secreted via SA is IL‐1β [39]. Indeed, 5xFAD mice retina showed intra‐ and extracellular accumulation of IL‐1β with immunohistochemical positivity in drusen‐like structures at the age of 12 months while the IL‐1β signal was absent in WT samples. Similarly, known cargos secreted via SA, HMGB1, vimentin, and ferritin [39, 41], were localized to basal lamina/Bruch's membrane at the age of 6 months, whereas WT animals showed only weak signals for vimentin. However, since the levels of vimentin and ferritin were analyzed only from 6‐month‐old 5xFAD samples, their possible accumulation to the drusen‐like structures seen mostly at the age of 12 months remained elusive.

The key intracellular players related to SA, TRIM16, galectin‐8, and Sec22B showed significant increase in 5xFAD retinas compared to WT mice. TRIM16 acts as a cargo receptor for SA, and it was found only from 5xFAD samples with increasing levels over time [39]. Also, TRIM16 seemed to be secreted from intracellular to extracellular space and accumulated to drusen‐like structures possibly along with the cargo destined for SA. Double staining with Aβ showed clear co‐localization in the drusen‐like deposits and partial co‐localization inside the RPE and the basal‐lamina/Bruch's membrane. However, the absence of TRIM16 signal in WT mice raises questions. The role of TRIM16 has been connected to the formation of protein aggregates, lysophagy, and control of NFE2L2‐Keap1 (nuclear factor erythroid 2‐related factor 2‐Kelch‐like ECH‐associated protein 1) signaling securing cellular oxidative stress homeostasis [55, 56]. Thus, TRIM16 seems to have a role in basal cellular functions. The absence of TRIM16 signal might be due to low expression at a basal level in the RPE cells and its recruitment to the autophagic cargo destined for secretory pathway enables its visualization in 5xFAD mice. However, the lack of functional analysis of TRIM16 in the SA and especially on Aβ secretion in this study remains elusive. Galectin‐8 and TRIM16 seem to cooperate in the induction process of SA deciding the secretion of the cargo [39]. 5xFAD mice showed significantly higher levels of galectin‐8 inside the RPE cells. Similarly, Sec22b connecting the cell membrane localized SNAP23/29 and Stx3/4 for delivery of the cargo outside from the cell [39] showed significant increase in the 5xFAD RPE cells. Galectin‐8 and Sec22b remained intracellular as expected due to their localization outside of the sequestration membrane and not along with the cargo as TRIM16. Also, the intracellular localization of Sec22b seemed to be more basal, especially at the three and six‐months' time points, pointing to the possible direction of the secreted material. At the age of 12 months, the more basal localization of Sec22b was lost and an intensive signal was seen throughout the cell. This change of Sec22b localization can be due to EMT where RPE cells lose their apical‐basal polarization. Subretinal drusenoid deposits (SDDs) located between the RPE and photoreceptor cells can also be seen in AMD patient retinas [36, 57, 58] suggesting a delivery of intracellular material to the apical side of RPE. However, the components of SDDs are distinct from basal drusen [58, 59, 60] possibly due to the localization of the aggregated material. Basal drusen contains several components originating from choroid and other sources [44], thus similar aggregation of material originating from neural retina can be anticipated in the case of SDDs. However, a recent publication connects the increase of SA intracellular players TRIM16, galectin‐8, and Sec22b and known SA cargo IL‐1β, vimentin, ferritin, and HMGB1 to the AMD patient population with drusen [61].

Many of the proteins secreted via SA are in connection with the EMT process [17, 18, 19]. HMGB1, vimentin, and ferritin levels were increased in the RPE cells of 5xFAD mice, suggesting EMT. Further analysis of collagen 1 and SNAIL levels in the RPE revealed their increase in the 5xFAD mice. The signal of SNAIL was absent, whereas low levels of collagen 1 were detected in the RPE cells of WT mice. Similarly, EMT markers are found to be upregulated in the AMD patient population [61]. Furthermore, the signal of RPE65 decreased in 5xFAD mice RPE cells at the age of 12 months, suggesting loss of differentiation [62]. These changes suggest the activation of EMT in 5xFAD mice.

Finally, the functional and morphological analysis of the retina and fundus revealed decreased visual functioning of the retina and accumulation of clearly visible drusen‐like material, respectively. However, only the scotopic responses in ERG analysis showed a clear decrease, and retinal thickness showed only a trend towards thinning despite the poor condition of the RPE cells and accumulation of drusen‐like material. However, being a nocturnal animal, mouse retina contains a high ratio of rod photoreceptors responsible for the low light vision [63] and, thus, may be more vulnerable to functional decline. Similar results have been seen previously when analyzing ERG, fundus, and ultrastructural images from 5xFAD mice [49, 64]. For example, the ERG recordings show significant changes at early time points, but the difference between 5xFAD and WT diminishes when older time points are analyzed. However, the use of a‐ and b‐waves in the ERG analysis cannot be linked to the RPE pathology. Recording of the c‐wave arising from the RPE cells would have narrowed the source of functional decline [65]. Therefore, the decrease seen in scotopic responses may arise from neural retina known to accumulate Aβ in the 5xFAD mice [34]. In OCT scans, the retinal thickness does not show clear differences towards thickening or thinning. Undigested POS, changes in Bruch's membrane, and vacuolization are common findings in ultrastructural analysis of the RPE. However, the increase in size of the vacuoles, especially autophagosomes, may suggest dysfunctional lysosomal degradation.

Despite overexpression and accumulation of Aβ in RPE cells and to the drusen‐like deposits, Aβ cannot be solely considered as the main factor of the disease phenotype nor formation of the drusen‐like deposits. Many cellular phenomena are known to contribute to the AMD‐like phenotype. For example, oxidative stress, autophagy dysfunction, inflammatory pathways (including the complement pathway), and EMT contribute to the disease development [10, 32, 66]. In the case of wet AMD, choroidal neovascularization and its regulators become crucial factors [67]. However, Aβ can indirectly affect these pathways and participate in disease development. For example, Aβ is known to cause oxidative stress, inflammatory response, and deficient lysosome/autolysosome acidification hampering the normal function of autophagy before extracellular deposits can be seen [68, 69, 70]. On the contrary, anti‐Aβ treatment has been shown to alleviate AMD‐like pathogenesis in model organisms [71, 72, 73]. In wet AMD, the formation of the new blood vessel from the choroid invading the retina can be an attempt to secure contact to the choroidal vasculature disturbed by the drusen accumulation between the choroid and the RPE cells [67]. Activation of SA can deliver pro‐angiogenic factors from RPE cells and contribute to the neovascularization [74]. Nevertheless, more precise analytics of Aβ overexpression/accumulation utilizing different techniques and models need to be done before the role of Aβ inducing AMD‐like pathogenesis including drusen‐like deposit formation can be established. Also, only female mice were used in this study to avoid single housing of male mice for a long period of time. This sets a sex‐specific limitation to the findings.

In conclusion, 5xFAD mice overexpressing Aβ showed accumulation of protein aggregates, including Aβ, beneath the RPE cells during aging. Closer inspection of these deposits revealed accumulation of markers usually connected to intracellular protein aggregation suggesting dysfunctional autophagic degradation. Furthermore, the shift of these accumulated markers to the extracellular space in drusen‐like structures beneath the RPE cells showed age dependency. Concurrently, markers of the SA pathway showed an increase and a similar shift from intracellular space to drusen‐like structures. Also, markers of EMT suggesting loss of epithelial cell characteristics were upregulated and oxidative stress was induced. These molecular findings became manifest as a formation of drusen‐like deposits clearly visible in the fundus. Thus, the 5xFAD mouse model holds many characteristics related to early AMD and can be considered as a model for drusen‐like deposit formation and secretory autophagy.

Author Contributions

Kai Kaarniranta and Ali Koskela: conceptualization. Kai Kaarniranta, Ali Koskela, and Elisa Toropainen: methodology. Johanna Ruuth, Elisa Toropainen, Juha M. T. Hyttinen, and Toni Tamminen: investigation. Johanna Ruuth and Toni Tamminen: visualization. Kai Kaarniranta, Johanna Ruuth, and Ali Koskela: funding acquisition. Kai Kaarniranta and Ali Koskela: supervision. Ali Koskela and Johanna Ruuth: writing – original draft. Kai Kaarniranta, Heikki Tanila, and Tarja Malm: writing – review and editing.

Funding

This work was supported by the Academy of Finland (Grant 333302), GeneCellNano, Paivikki ja Sakari Sohlberg Foundation, Kuopio University Hospital district (Grant 5503770), Sigrid Juséliuksen Säätiö (Sigrid Jusélius Stiftelse), Silmäsäätiö (Finnish Eye Foundation), De Blindas Vänner (DBV), Finnish eye and Tissue Bank Foundation, Mary and Georg C. Ehrnrooths Foundation, and Finnish Cultural Foundation‐North‐Savo.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: The Rd8 mutation detection with PCR. The possible presence of the Rd8 mutation in the Crb1 gene of NRF‐2/PGC‐1α KO mice was detected with PCR amplification [3]. Three primers were used: forward primers mCrb1 mF1, 5′‐GTGAAGACAGCTACAGTTCTGATC and mCrb1 mF2, 5′‐GCCCCTGTTTGCATGGAGGAAACTTGGAAGACAGCTACAGTTCTTCTG, and a reverse primer mCrb1 mR, 5′‐GCCCCATTTGCACACTGATGAC [3]. For PCR amplification, 3 μL of genomic DNA isolated from the mouse ear samples were used in a 25‐μL reaction volume, containing reaction buffer, 1.5 mM MgCl₂, 100 μM of each dNTPs, 1.6 μM each of the forward primer for WT allele (F1), 0.8 μM of forward Rd8 mutant allele (F2), and 1.6 μM of reverse primer, and 1.2 U of DreamTaq DNA polymerase. Samples were denatured at 95°C for 2 min, followed by 34 cycles at 95°C for 30 s, at 65°C for 30 s, and at 72°C for 30 s, and a final extension at 72°C for 7 min. PCR products were separated in 2.5% agarose gel electrophoresis with ethidium bromide stain, and photographed on UV. The fragment size for the WT allele is 220 bp and for the mutant Rd8 allele – 244 bp, respectively. S1‐S5 samples; Rd8+, Rd8 positive control; MW, molecular weight; bp, base pairs.

Figure S2: Double staining of APP‐CTF (Sigma (A8717) Anti‐Amyloid Precursor Protein, C‐Terminal antibody) and 22C11 (Merck, MAB348 Anti‐APP A4 Antibody, a.a. 66‐81 of APP {NT}, clone 22C11). Signals colocalized in RPE layer of WT mice. In 5xFAD mice there was no signal of APP‐CTF or 22C11 in RPE layer. When the signal of 22C11 is positive but the signal of APP‐CTF is negative, the stained molecules are sAPPα or sAPPβ. In WT both APP‐CTF α and β were positive, the signal is full‐length APP. Therefore, the signal detected in 5xFAD is 42 amino acids long amyloid beta.

Figure S3: Use of autophagy reporter model CAG‐RFP‐EGFP‐LC3 combined with 5xFAD showed more autophagic structures than CAG‐RFP‐EGFP‐LC3 alone. (A) 5xFAD strain was combined with CAG‐RFP‐EGFP‐LC3 reporter mice to visualize autophagic structures. CAG‐RFP‐EGFP‐LC3 transgenic mice have CAG promoter/enhancer sequences driving expression of a red fluorescent protein (RFP), an enhanced green fluorescent protein (EGFP), and a microtubule‐associated protein 1 light chain 3 alpha (Map1lc3a or LC3) gene. The dual expression of two fluorophores allows tracking of different types of phagocytic cellular compartments depending on the acidity of the environment. RFP is stable in acidic pH (pK _a  4.5) while EGFP (pK _a  5.9) is quenched in the acidic lysosomal environment, allowing autophagosomes to be distinguished from autolysosomes. The genotyping of CAG‐RFP‐EGFP‐LC3 reporter mice was performed with a 4‐primer PCR. The primer sequences were: GFP‐5′; 5′‐CAT GGA CGA GCT GTA CAA GT, LC3B‐3′; 5′‐CAC CGT GAT CAG GTA CAA GGA, IPC‐5′; 5′‐CTA GGC CAC AGA ATT GAA AGA TCT, and IPC‐3′; 5′‐GTA GGT GGA AAT TCT AGC ATC ATC C. Reaction mixtures included 100 nM of each primer, 100 nm of each dNTP, 4 μL of the previously prepared mouse DNA, and 0.16 μL of DreamTaq DNA polymerase (5U/μL; ThermoFisher), an a total volume of 20 μL. Reaction profile was: initial heating at +95°C for 1 min, followed by 32 cycles consisting of at +95°C for 30 s, at +50°C for 30 s, and at +72°C for 30 s, ensued by the final extension at +72°C for 7 min. DNA products were separated using 1.7% agarose gel electrophoresis and visualized with EtBr and UV light. Exposing the transgenic sample, both 208‐bp and 324‐bp fragments were seen, and for wild type sample only the 324‐bp fragment produced by the IPC primers was detected. The mice were euthanized and eyes processed as written in the main text. Sections were cut and slides were sealed with mounting media and imaged as soon as possible. (B) Original images of 6 months old CAG‐RFP‐EGFP‐LC3 reporter and CAG‐RFP‐EGFP‐LC3/5xFAD mice RPE cells. (C) Signal threshold adjusted 8‐bit images of CAG‐RFP‐EGFP‐LC3 reporter and CAG‐RFP‐EGFP‐LC3/5xFAD mice RPE cells showed lipofuscin and phagosome punctate. Lipofuscin punctate shows both red and green autofluorescence, 5xFAD mice RPE cells are known to accumulate lipofuscin particles. CAG‐RFP‐EGFP‐LC3 reporter mice are not expected to accumulate lipofuscin to the RPE cells. (D) Lipofuscin particles were removed from the 8‐bit images using filtering. According to [75], lipofuscin punctate can be filtered out with ImageJ Analyze particles feature by using size and shape in red channel. Since CAG‐RFP‐EGFP‐LC3 reporter mice are not expected to accumulate lipofuscin, these punctate can be assumed as a true signal. The filtering settings for size and circularity were determined using CAG‐RFP‐EGFP‐LC3 reporter mice punctate and applied to CAG‐RFP‐EGFP‐LC3/5xFAD model. The values for size were set to 0.1–30.00 px² and 0.05–1.00 for circularity for both mice strains. The CAG‐RFP‐EGFP‐LC3 reporter mice showed only few autophagosome structures (red and green signal co‐localization) whereas to CAG‐RFP‐EGFP‐LC3/5xFAD mice showed exclusively autolysosomes (only red signal). Also, the number of red punctate was higher in CAG‐RFP‐EGFP‐LC3/5xFAD mice. Autophagy flux and the turnover of autophagic structures is rapid and therefore the high number of autolysosomes detected in CAG‐RFP‐EGFP‐LC3/5xFAD mice suggests delayed degradation. Simultaneously, no new autophagosomes are formed suggesting dysfunctional autophagy flux. This is in line with Figure 1C where no autophagy cargo receptor SQSTM1 signal can be found inside the RPE cells. However, detection of autophagosomes from RPE cells using CAG‐RFP‐EGFP‐LC3 reporter mice model is challenging and may result in overestimation of autophagosome size [75]. WT = wild type; 5xFAD = five timed familial Alzheimer's disease mutation; RFP = red fluorescent protein; EGFP = enhanced green fluorescent protein; LC3 = microtubule‐associated protein 1 light chain 3 alpha; CAG = CMW early enhancer/Chicken beta actin. Scale bar = 10 μm.

Figure S4: Oxidative stress marker HNE and antioxidant enzyme catalase levels were elevated in the RPE cells of 5xFAD mice at the age of 12 months. (A) The level of HNE was significantly lower in the RPE cells of 5xFAD mice at the age of 6 months. As the mice aged, the level of HNE increased significantly in the 5xFAD mice RPE cells. (B) The antioxidant enzyme catalase level was significantly lower in the RPE cells of 5xFAD mice at 6 months of age. The catalase level increased significantly in the 5xFAD mice at 12 months of age. The catalase signal was present beneath the Bruch's membrane in WT and 5xFAD mice at 12 months of age. The localization of catalase signal in the choroid was more prominent in the 5xFAD mice (white circle). Bruch´s membrane (white arrow), RPE cells (yellow arrow). Scale bar 10 μm, *p < 0.05, ****p < 0.001, Mann‐Whitney U‐test.

Data Availability Statement

The data that support this study are available on request from the corresponding author.