Animal models

A total of 63 female TAU58 heterozygous transgenic mice were enrolled in the study: 30 underwent subretinal injections, while 33 remained non-injected, across various age groups (6 months, n = 9; 9 months, n = 10; 12 months, n = 10; 15 months, n = 4). Furthermore, 14 wild-type (WT) littermates (6 months, n = 7; 12 months, n = 6) were included for comparative analysis. TAU58 transgenic mice were specifically chosen for this study because they express human 4-repeat Tau with the P301S mutation, driven by the Thy1 promoter [41]. Notably, TAU58 mice begin developing cortical and brainstem pTau lesions at an early stage, as early as 2 months of age, with a progressive increase in quantity and distribution across various brain regions with advancing age [41]. The transgenic expression of pTau ensures the presence of sufficient pTau in the retina to induce/accelerate its aggregation by seeds [41, 46]. Previous studies have further confirmed the expression of pTau in the retinas of mice carrying the P301S mutation [47, 48].

Female mice were used for these experiments to reduce sex-related variations in the results leading to a reduction in the number of animals in accordance with the 3R principle [49]. Previous studies have not reported mechanistic differences between male and female mice regarding pTau seeding and propagation [34, 35, 50, 51], supporting our rationale for restricting the study to female mice. Studies showed that female TAU58 mice have a milder pTau pathology profile, offering distinct advantages in controlled seeding effects within the superior colliculus (SC), lateral geniculate nucleus (LGN), and the primary visual cortex (V1) [41, 43].

TAU58 heterozygous mice were crossbred with C57BL/6J WT mice. Retinas were examined histopathologically, and any mice exhibiting visible ocular abnormalities were excluded from the study.

Genotyping was performed as reported [41]. Mice were maintained on a 12-hour light/dark cycle, with access to food and water ad libitum. All animal care and experimental procedures were conducted in accordance with the ethical standards set by the KU Leuven Ethical Committee (P169/2020) and in compliance with Belgian and European law.

Human autopsy cases

Human brain autopsy brain tissue (temporal cortex Brodmann area 21/22) form two patients was used in this study to prepare seeds with negligible and high pTau levels: 1 non-AD control with PART (Braak stage I) and 1 AD patient (Braak Stage VI) (Tab. S1). The PART case showed Braak stage I. Accordingly, the temporal cortex of Brodmann areas 21/22 of the PART case did not exhibit microscopic pTau pathology. Given that nearly all individuals over the age of 40 years develop at least initial pTau lesions in the brain, this non-AD case with an initial PART stage is considered representative of the age-matched general population [18]. Therefore, it was used as a “non-disease” control in this study. The brains were collected at UZ Leuven (Belgium) and the laboratory of Neuropathology of Ulm University (ethical approval: 54/08 (Ulm); S-52971 (Leuven)) and used for this study under approval by the ethical committee from UZ Leuven (S-64492) and in accordance with the Belgian law. Neuropathological characterization of all cases was performed by D.R. Thal.

The phases of Aβ deposition within the medial temporal lobe (referred to as AβMTL phases) were evaluated according to recommended criteria [52, 53], serving as a valid approximation for Aβ accumulation throughout the entire brain [16]. The progression of NFTs across the brain was determined using Braak NFT stages, utilizing sections immunostained with anti-pTauS202/pT205 (AT8, Invitrogen Thermo Fisher Scientific, 1/1000) [32]. Specifically, Braak NFT stages were assessed in accordance with the methodology established by Braak et al. [18].

The frequency of pTau-positive neuritic plaques was evaluated following the guidelines set by the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) [54]. Furthermore, the degree of AD pathology was determined according to the National Institute of Aging-Alzheimer Association (NIA-AA) [52]. This determination took into consideration the Aβ-phases, the Braak NFT stage, and the CERAD score pertaining to neuritic plaque pathology. The distribution of TDP-43 pathology was determined according to the recommendations for the neuropathological assessment of limbic predominant age-related TDP-43 encephalopathy neuropathological changes (LATE-NC) [55].

Protein extraction and BCA assay

To extract insoluble material containing Tau protein, fresh-frozen brain tissue from the temporal cortex underwent a series of centrifugation steps using progressively stronger detergents, following a previously established protocol [56]. Initially, the brain tissue was homogenized in a buffer consisting of high salt and Triton X (HS-TX) and then subjected to ultracentrifugation at 121,656 × g for 30 min at 4 °C. The resulting pellet was subsequently resuspended in HS-TX solution containing 20% sucrose, followed by another round of ultracentrifugation at 121,656 × g for 30 min at 4 °C. Next, the resulting pellet was resuspended in HS-TX solution containing 2% sarkosyl and subjected to ultracentrifugation at 121,656 × g for 30 min at room temperature. Afterwards, the pellet was resuspended in phosphate-buffered saline (PBS) and underwent two rounds of ultracentrifugation at 121,656 × g for 30 min each, at room temperature. This step was performed to eliminate the detergents from the homogenates. The final supernatant was resuspended in PBS, briefly sonicated, and then stored at -80 °C for further use. The overall protein content was measured utilizing the Pierce® BCA Protein Assay Kit (ThermoFisher Scientific) according to the manufacturer’s instructions.

ELISA

To determine the concentration of pTau in patient-derived brain homogenates, an ELISA assay was conducted (pTau181 human ELISA Kit, ThermoFisher Scientific, #KHO0631), following the manufacturer’s instructions. Duplicates were run for all samples, and final concentrations were calculated by interpolating within the polynomial range of the standard curve for each assay. The resulting interpolated concentrations and the specific amounts of injected pTau per site are presented in Table S1.

Tau biosensor cell line

To confirm the seeding potential of the sarkosyl-insoluble fraction derived from the AD and non-AD/PART cases, we used the Tau repeat domain (RD) P301S FRET Biosensor HEK-293 cell line (CRL-3275, ATCC- LGC Standards, Molsheim Cedex, France), which stably expresses constructs where the RD of Tau-P301S is fused to CFP and YFP. Incubation with Tau seeds nucleates the aggregation of these Tau reporter proteins, thereby producing a FRET signal [57]. For seeding experiments, the cells were cultured in DMEM medium, supplemented with 10% FBS, 1 mM sodium pyruvate, and non-essential amino acids (Gibco, Thermofisher Scientific, Waltham, MA, USA), under an atmosphere of 5% CO2 at 37 °C. Cells were plated at 5.000 cells/well in poly-L-Lysine-coated 384-well PhenoPlates (PerkinElmer, Mechelen, Belgium). After 16 h, cells were incubated with brain extracts using Lipofectamine 3000 (Thermofisher Scientific, Waltham, MA, USA) according to the manufacturer’s protocol. Before the incubation, the samples were sonicated for 15 min (30 s on, 30 s off at 10 A) (Bioruptor Pico, Diagenode, Seraing, Belgium). Each sample was mixed with Lipofectamine 3000 reagent and added to a mixture of Opti-MEM medium (Gibco, Thermofisher Scientific, Waltham, MA, USA) with Lipofectamine 3000. After a 15 min incubation at room temperature, 4µL of mixture was added per well in a total volume of 40µL. After 48 h, cell medium was replaced with 40µL 4% formaldehyde for 5 min, then washed three times with PBS. Nuclear staining was performed with DAPI (Thermofisher Scientific, Waltham, MA, USA). Three individual plate preparations were performed per sample as independent experiments (n = 3). High-content imaging was performed at the VIB Imaging Core (Leuven, Belgium) (Operetta, PerkinElmer, Mechelen, Belgium). Segmentation analysis of 15 fields in 6 planes at a 40x magnification was performed using the Columbus Plus digital platform (PerkinElmer, Mechelen, Belgium).

Subretinal injections

Two-month-old female TAU58 mice (Tab. S2) were anesthetized by intraperitoneal (i.p.) injection of a mixture of ketamine (Nimatek, 100 mg/mL) and medetomidine (Domitor, 1 mg/mL) in saline. In addition, eye drops with local anesthesia (oxybuprocaine 0.4%, Unicaïne, Théa Pharma) were applied on the injected eye, followed by pupil dilation eye drops (tropicamide 0.5%, Tropicol, Théa Pharma). Carbomer hydrogel (carbomer 0.2%, Vidisic, Bauch & Lomb) was applied on the non-injected eye to prevent drying of the cornea. The subretinal injection was performed under direct visualization of the fundus. A trans-vitreal injection of 1 µL of the sarkosyl insoluble fraction of the human brain lysates into the subretinal space of the lateral part of the right eye was performed using a Hamilton syringe (65RN, 600, Fisher 7633-01) equipped with a beveled 34G needle (6x, GA 33, 51 mm, Fisher 7762-06). The injections were performed with standardized lysate volume representing the sarkosyl insoluble fraction of approx. 0,5 mg brain tissue/µl. The pTau concentration in the lysates was determined by ELISA as described. Following the injection, antibiotic treatment (Tobrex, tobramycinum 3 mg/g, Alcon) was applied to the injected eye. The animals were sacrificed 16 weeks later at 6 months of age by decapitation under terminal anesthesia. Eyes and brains were removed for pathological and immunohistochemical analysis and fixed in 4% phosphate-buffered paraformaldehyde.

Immunohistochemistry

The mouse brains and eyes were fixed in 4% paraformaldehyde for three days and then embedded in paraffin. Paraffin-embedded tissue sections were cut at 5 μm thickness and subsequently deparaffinized. Heat-induced epitope retrieval was carried out using Envision™ Flex Target Retrieval Solution Low pH (Dako, K8005) for 10 min at 97 °C. Endogenous mouse peroxidase was blocked with a Peroxidase-Blocking Reagent (Dako) for 5 min to prevent nonspecific reactions. Mouse anti-pTauS202/T205 primary antibody (AT8, Invitrogen Thermo Fisher Scientific, 1/500) conjugated to biotin was applied and incubated overnight. Immunohistochemical labeling was conducted using the ABC method, as previously described [58]. 3,3’-diaminobenzidine was utilized as a chromogen, and counterstaining was performed with hematoxylin using a Leica Autostainer (Leica, Wetzlar, Germany).

Fluorescence staining was conducted using overnight incubation with the following primary antibodies: polyclonal guinea pig anti-NeuN (neuronal nuclei, Synaptic Systems, 1/200), rabbit anti-RBPMS (RNA-binding protein with multiple splicing, PhosphoSolutions, 1/250), goat anti-Iba1 (ionized calcium-binding adapter molecule 1, Abcam, 1/300), guinea pig anti-GFAP (glial fibrillary acidic protein, Synaptic Systems, 1/400) and biotin-conjugated mouse anti-pTauS202/T205 antibody (AT8, Invitrogen Thermo Fisher Scientific, 1/500). Subsequently, a 90-minute incubation was performed with the secondary antibodies: goat-anti rabbit (conjugated with Cy2 fluorochrome), donkey-anti goat (conjugated with Cy2 fluorochrome), goat-anti guinea pig (conjugated with Cy5 fluorochrome), and streptavidin (conjugated with Cy3 fluorochrome). Hoechst 33,342 stain (Thermo Fisher Scientific) was used to visualize nuclei. Slides were mounted using a Glycergel mounting medium (Dako). Microscopic images of the eyes were captured using a DCC 290 microscope paired with a 20 × or 40 × objective with a DFC7000 T camera (Leica). Mouse brain slides were scanned using the Zeiss Axio Scan Z.1 (Zeiss) at a 10× magnification (0.65 μm/pixel).

Positive and negative staining controls were performed.

Gallyas silver staining and quantification of fibrillar tau pathology

The silver-impregnation method of Gallyas-Braak was used to visualize filamentous Tau pathology on paraffin-embedded, 5 μm coronal sections, as previously described [59, 60]. Both positive and negative staining controls were included to ensure staining specificity.

For the quantification of fibrillar Tau, Gallyas staining was performed on both the subretinally injected (right) eyes and the non-injected (left) eyes across the examined groups (PBS, non-AD/PART, AD). Five retinal sections were analyzed: one central section containing the optic nerve, with each of the remaining sections spaced 150 μm apart, dorsally and ventrally. Only sporadic fibrillar material was seen in the retina. To simplify the interpretation the examination of the changes was performed on the entire retina cross-sections without separation into the peripheral and central parts. In the optic projection areas (SC, LGN, V1), four sections per brain region, spaced 250 μm apart, were analyzed to cover the entire regions of the SC, LGN, and V1, separately for the respective ipsi- and contra-lateral hemispheres. The presence/absence of fibrillar aggregates was presented as binary data for clarity and ease of interpretation (Tab. S3). Image analysis was performed in a blinded manner.

Quantification of neuronal density and pTau pathology

After immunolabeling with anti-NeuN and anti-pTauS202/T205 antibodies, brain sections were scanned (Zeiss Axio Scan Z.1) and images were imported into QuPath image analysis software [61]. Three sections per brain, spaced 250 μm apart, were analyzed to cover the layers of the LGN, SC, and V1. Neurons were selected from the NeuN channel, using a defined threshold for precise visualization, and cell validation was performed through colocalization with Hoechst 33,342 nuclear staining. Neuronal density (NeuN-positive neurons/mm²) and the ratios of pTau-positive (NeuN-positive, pTauS202/T205-positive) neurons relative to the total number of neurons (NeuN-positive) were quantified. Specific layers of the LGN, SC, and V1 were manually outlined (Fig. S1), referencing the Allen Reference Atlas – Mouse Brain (atlas.brain-map.org).

For mouse retinas, three sections were analyzed, including the central section with the optic nerve, and two sections extending 150 μm dorsally and ventrally, respectively (Fig. S2). QuPath was used for manual quantification of RBPMS-positive retinal ganglion cells and the pTauS202/T205 immuno-positive area across the retinal ganglion cell layer. Per section, the number of RBPMS-positive cells was quantified in four regions of interest, comprising both the central and peripheral retina (Fig. S2a). The rationale for using the RBPMS marker was to distinguish the retinal ganglion cells from amacrine and glial cells [62, 63]. The presence of pTau was measured by assessing the total length of the pTauS202/T205 positive area versus the total length of the retina, and the results were expressed as a percentage (Fig. S2b).

The evaluation of pTau threads in the optic nerve of mice receiving subretinal injections was conducted on one central sagittal section of the eye, on which the optic nerve was visible. The presence or absence of pTau alterations in each optic nerve (left/right) was evaluated via immunostaining with the anti-pTauS202/T205 antibody, and presented as binary data for clarity and simplicity of interpretation.

Image analysis was performed in a blinded manner.

Quantification of the microglia and astrocytes

A morphometric quantitative analysis was performed to measure the percentage of GFAP-positive area, and percentage of Iba1-positive area, in the right (subretinally injected) eyes across the groups (PBS, non-AD/PART, AD) and the corresponding contralateral brain regions (SC, LGN, and V1) using specific antibodies to detect astrocytes (GFAP) and microglia cells (Iba1) following immunofluorescence staining. Digital images of the peripheral and central retinal regions, as well as two images per brain region (SC, LGN, and V1), covering the entirety of each region, were captured using a Leica DM2000 LED microscope equipped with a Leica DFC 700T camera at ×10 magnification and processed using LAS V4.8 software.

For each mouse, four images were taken from the peripheral retina (defined as 100 μm from the ora serrata) and four from the central retina (defined as 100 μm from the optic nerve). Additionally, two images were taken from each contralateral brain region (SC, LGN, and V1), analyzing the whole region for each.

The percentage of GFAP- and Iba1-positive areas were measured using ImageJ software (NIH, Bethesda, USA). For the retina, measurements were taken over a 1 mm length in both peripheral and central regions, encompassing all retinal layers except the photoreceptor layer. For the brain regions (SC, LGN, V1), the entire region imaged and analyzed at the same magnification to ensure consistency. The presence of astrocytes and microglia was quantified as the percentage of the GFAP- or Iba1-positive area relative to the total ROI area. A brightness threshold was applied to detect GFAP-positive astrocytes and Iba1-positive microglia, with the consistent threshold settings in ImageJ ranging between 64 and 255 for GFAP and 90–255 for Iba1. All samples were stained simultaneously on the same bench to ensure uniformity, and background noise was carefully eliminated to include only GFAP- and Iba1-positive areas in the analysis. Image analysis was performed in a blinded manner.

Statistical analysis

Statistical analyses were performed using Prism v.9 (GraphPad) and IBM-SPSS Statistics 25 (SPSS, Chicago, IL, USA). Linear regression investigated the association between pTau-positive area percentage and age in non-injected TAU58 mice across different ages (Tab. S4). Two-way ANOVA with Tukey’s test compared RBPMS-positive ganglion cells between 6-month-old and 12-month-old TAU58 mice and age-matched WT littermates, separately for peripheral and central retina (Tab. S5). Linear regression analyzed the association between RBPMS-positive ganglion cells and age, separately for peripheral and central retina (Tab. S6). Linear regression further examined the association between RBPMS-positive ganglion cells and pTau-positive area percentage, with age as a controlled variable, separately for the peripheral and central retina (Tab. S7). One-way ANOVA with Tukey’s test assessed differences in pTau seeding potential among brain homogenates (Tab. S8). One-way ANOVA with Dunnett’s test compared pTau-positive area percentage after subretinal injection with brain lysates, with separate analyses for left and right eyes, using PBS groups as a control (Tab. S9). An unpaired t-test compared pTau-positive area and RBPMS-positive neurons between PBS-injected and non-injected eyes (Tab. S10). Multinomial logistic regression evaluated the association between fibrillar Tau presence in the right injected eye and the respective injected brain lysates, using the PBS group as the reference category (Tab. S11). Multinomial logistic regression evaluated the association between pTau threads presence in the optic nerve and injected brain lysates, separately for the left non-injected eye and the right injected eye, using the PBS group as the reference category (Tab. S12). A one-way ANOVA with Dunnett’s test compared RBPMS-positive neurons after subretinal injections with brain lysates, with separate analyses for the peripheral and central regions of the left eye and the peripheral and central regions of the right eye, using PBS groups as a control (Tab. S13). A one-way ANOVA with Dunnett’s test compared the percentage of GFAP and Iba1-positive areas in the central and peripheral retina of injected eyes, using PBS as control (Tab. S14). Linear regression examined the association between the percentage of the pTau-positive area in the right (injected) eye and the percentages of GFAP-positive and Iba1-positive areas, separately for the peripheral and central retina (Tab. S15). Kruskal-Wallis test (non-parametric one-way ANOVA) with Dunn’s test assessed pTau pathology percentage in brain regions (SC, LGN, V1) after subretinal injection with brain lysates, with separate analyses for each brain layer, using PBS groups as a control (Tab. S16). A one-way ANOVA with Dunnett’s test examined NeuN-positive neuron count in brain regions (SC, LGN, V1) after subretinal injection with brain lysates, with separate analyses for each brain layer, using PBS groups as a control (Tab. S17). Finally, a one-way ANOVA with Dunnett’s test compared the percentage of GFAP and Iba1-positive areas in contralateral optic areas (SC, LGN, V1), between injected lysates, with separate analyses for each brain, using PBS as control (Tab. S18). Differences were considered statistically significant for two-sided p-values < 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).