Design and application of a Tau seed amplification assay for screening inhibitors of Tau seeding

Abstract

Background

Tau protein aggregates are a key pathological hallmark of Alzheimer’s disease (AD) and are closely associated with cognitive decline and neurodegeneration. It is proposed that tau aggregates faithfully propagate throughout the brain by self-templating their disease-associated conformation onto natively-folded tau monomers, thereby inducing their aggregation and incorporation into growing fibrils. As such, the inhibition or modulation of tau seeding and aggregation represents a viable therapeutic strategy for AD and other tauopathies.

Methods

We have recently developed seed amplification assays (SAA) for the detection and amplification of small quantities of misfolded protein aggregates in various neurodegenerative diseases. In this article, we adapted the SAA technology to amplify the process of tau aggregation and seeding in AD brain samples. Using the Tau-SAA we screened two chemical libraries: one comprising over 20 suspected aggregation inhibitors and the other comprising over 200 FDA-approved, blood-brain barrier-permeable compounds from a commercial chemical library. We also performed secondary in vitro assays to confirm the activity of selected hits as well as determining the IC50 of the most active compounds.

Results

Our Tau-SAA detects the presence of tau seeds even after a 100-million-fold dilution of the initial inoculum. Examination of 26 postmortem brain samples from AD and control cases confirmed that our assay is specific for AD brain tau seeds. Screening of 220 compounds showed that approximately 57% of suspected aggregation inhibitors and ~ 3% of CNS-penetrant compounds inhibited over 75% of AD brain-templated tau aggregation.

Conclusions

In conclusion, our data suggests that Tau-SAA readily detects the presence of tau seeds in AD brains but not in controls, and that by amplifying AD brain tau seeds, the assay may serve as a valuable primary drug screening platform.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-025-01855-y.

Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder associated with gradual memory loss and deterioration of executive function [1]. Neuropathologically, the disease is characterized by the accumulation of extracellular amyloid plaques, primarily composed of amyloid-beta (Aβ) and intracellular neurofibrillary tangles composed of hyperphosphorylated tau [2, 3]. Despite extensive efforts to develop disease-modifying therapies, AD remains a largely untreatable and incurable disease [4].

Recent therapeutic strategies primarily targeted Aβ in hopes of slowing disease progression. However, the efficacy of these Aβ-directed approaches remains debatable and has been the subject of considerable controversy [5, 6, 7]. Aducanumab, the first Aβ immunotherapy to receive accelerated approval from the Food and Drug Administration (FDA), was widely criticized due to concerns over its limited clinical benefit and potential risks, ultimately leading to its discontinuation in 2024 [8]. Another Aβ-targeting drug, lecanemab, was granted full approval by the FDA [9] but was later rejected by the European Medicines Agency (EMA) due to differing interpretations of clinical trial results and concerns regarding safety and efficacy [10]. Most recently, a third candidate, donanemab, also received full FDA approval, further fueling the debate over the clinical value and risk associated with these anti-Aβ drugs [5, 10, 11, 12, 13]. Despite these ongoing controversies, the thus far limited success of Aβ-targeting therapies underscores the need for alternative approaches. Mounting evidence suggests that tau pathology more closely correlates with the loss of cognitive function, warranting a more thorough investigation of tau-directed therapies [14, 15, 16, 17].

A key event in AD and other tauopathies is the progressive spreading of the pathology to different brain regions, which is likely the cause of the progressive and chronic degeneration observed in these diseases. The propagation of tau pathology in AD and other tauopathies is thought to occur via a “prion-like” seeding-nucleation mechanism, in which pathological tau triggers the misfolding and aggregation of natively folded tau monomers into the disease-associated form [18, 19, 20, 21]. Tau seeding has been convincingly demonstrated in cellular [22, 23, 24] and animal models [25, 26] of tau pathology. Moreover, the prion-like cell-to-cell transmission of tau through anatomically connected brain areas [27, 28] provides a molecular explanation for the stereotypical spread of tau pathology in the human brain [29]. Neuropathological studies have shown that tau pathology starts in a rather small area of the brain, much before the appearance of clinical symptoms. The pathology progressively extends to new brain areas during the disease progression, increasing the extent of brain damage [29]. Therefore, the main driver of neurodegeneration appears to be the expansion of tau pathology in the brain, and thus, blocking tau spreading represents a great therapeutic target.

Various tau-directed strategies have been considered for possible therapeutic intervention including the reduction of tau expression, targeting post-translational modifications, stabilizing microtubules, immunotherapy, and tau aggregation inhibitors [30, 31]. Several small molecules and natural compounds have been shown to inhibit tau aggregation [31], but only three have been tested in clinical trials [30]. A phase II clinical study (NCT01383161) investigating the effect of curcumin, a bioactive component found in turmeric, noted that patients receiving the drug exhibited improvements in attention, long-term retrieval, and visual memory along with decreased tau and Aβ burden by positron emission tomography (PET) ligand binding [32]. Phase III trials of another compound, the methylene blue derivative LMTX, did not reach primary efficacy outcomes [33], although later reports claimed the treatment was effective in a smaller subset of enrolled patients [34]. While these early trials illustrate the challenges of developing clinically effective therapies, they also affirm the potential of targeting tau pathology as a therapeutic strategy. As interest in tau-directed therapies continues to grow, there is a critical need for robust, sensitive platforms to evaluate and identify compounds that effectively inhibit tau aggregation and seeding.

Fluorescence-based aggregation assays have been readily utilized to identify tau inhibitors [35, 36]. These in vitro assays test the potential of various compounds to slow or nullify the spontaneous aggregation of recombinant human tau proteins or peptides. Traditionally, these assays have relied on exogenous anionic inducers, high temperatures, and constant agitation to induce the formation of tau aggregates [36]. While such conditions allow for rapid aggregation assays, recent high-resolution structural studies have demonstrated that these spontaneously generated tau filaments are conformationally heterogeneous and distinct from those derived from the AD brain [37, 38]. These findings have raised doubt about the biological relevance of spontaneously generated fibrils and sparked interest in the use of brain-derived or disease-relevant fibrils for tau aggregation inhibitor discovery [39].

Here, we present the design and application of a tau seed amplification assay (SAA) as an improved tau inhibitor screening platform. The SAA (also known as protein misfolding cyclic amplification or PMCA and real-time quaking induced conversion or RT-QuIC) technology reproduces the prion-like seeding mechanism responsible for the spreading of protein aggregates in a cell-free system [21, 40, 41]. In SAA, prion-like ‘seeds’ from a biological sample are replicated at the expense of a recombinant protein substrate through cycles of shaking and incubation. Amplification can be monitored in real-time by fluorescence of the amyloid binding dye, Thioflavin T (ThT). Extensive studies have already demonstrated that SAA can faithfully propagate strain-specific biological properties in prion diseases [40, 42], and recent studies suggest similar outcomes in SAA-generated α-synuclein fibrils [43, 44].

While several groups have established Tau-SAA for the detection of AD and tauopathy brain-derived tau seeds [45, 46, 47, 48, 49, 50, 51, 52, 53], the structural fidelity of amplified fibrils remains an open question. Encouragingly, recent Cryo-EM studies of tau protofilaments extracted from cells seeded with brain-derived tau aggregates have shown that the resulting fibrils retain core structural features of the original seed [54], suggesting that tau fibrils generated through seeded aggregation can maintain disease-relevant features. The potential of Tau-SAA to similarly retain key structural characteristics of AD-derived tau aggregates enhances its utility as a model system for investigating aggregation inhibitors in a disease-relevant context. Moreover, Tau-SAA models specifically the seeding process, responsible for the spreading of tau pathology in the brain, possibly the best target for AD treatment.

In the present study, we first determined the tau isoform most suitable for Tau-SAA, and then validated our assay by examining the seeding activity of 26 postmortem brain samples from AD and control cases (HC). To begin to characterize the structural properties of the amplified fibrils, we evaluated their proteolytic resistance and analyzed the amino acid composition of the protease-resistant core. Finally, we utilized Tau-SAA as a drug screening platform by testing the inhibitory potential of over two hundred compounds from two different chemical libraries. Our findings support the use of Tau-SAA not only as a sensitive detection assay but also its potential as a valuable tool for identifying and developing tau aggregation inhibitors against disease-relevant AD-brain templated fibrils.

Methods

Expression and purification of Tau isoforms

All six human CNS tau isoform constructs were designed, expressed and purified in a similar manner. Briefly, sequences for full-length tau isoforms were downloaded from the Uniprot database (P10636). To create cysteine-free constructs, cysteine residues were substituted for serine. These modifications were made to preclude oligomerization through cysteine bridges. Two stop codons were added at the C-terminal end. Cloning cassettes were synthesized and cloned into the bacterial expression vector pET-28b using NcoI and XhoI restriction sites by GenScript. Other than aforementioned cysteine to serine point mutations, all constructs retain the canonical human tau sequences and are free of any protein tags or foreign cleavage sites.

Plasmids for all constructs were transformed into BL21(DE3) Escherichia coli and cultured using an auto-induction expression protocol. Briefly, 25 µl of a previously prepared glycerol stock were inoculated into a starter culture of 25 ml of LB-Miller media with 30 µg/ml kanamycin and incubated at 37 °C with 225 rpm agitation. After 8 h, the starter culture was diluted 1:100 into 500 ml of Terrific Broth auto-induction media (Formedium) with 30 µg/ml kanamycin and incubated at 37 °C with 225 rpm agitation overnight. The next morning, the bacteria were harvested by 3000 x g centrifugation at 4 °C for 30 min. The supernatant was discarded and bacterial pellets were collected and stored at -20 °C until use.

For the purification of tau protein, bacterial pellets were first resuspended in 20 mM PIPES pH 6.5 with 500 mM NaCl. Resuspended bacteria were lysed with 3–4 rounds of probe sonication on ice (Misonix 3000), where each round consisted of 5 min, 10 s on, 10 s off at a power of 6.5. Lysates were then boiled for 20 min at 95 °C, cooled on ice for 30 min, and centrifuged (15,000 x g) for 20 min at 4 °C. Supernatants were collected and centrifuged again. The resulting supernatant was pooled and ammonium sulfate was added to 55% (w/v). The mixture was allowed to incubate for 1 h with constant stirring after which precipitated proteins were collected by centrifugation at 15,000 x g for 15 min. The supernatant was discarded and pellet stored at -20 °C until use.

Tau proteins were purified using cation exchange chromatography on an ÄKTA pure chromatography system (Cytiva). The aforementioned pellet was resuspended in 20 mM PIPES pH 7.4 until the conductivity fell below that of Buffer A (20 mM PIPES pH 6.5 50 mM NaCl), typically ~ 9 mS. The resultant solution was passed through 0.2 μm filter and loaded onto two 5 ml HP SP columns (Cytiva) connected in series. After sample application, the column was washed with Buffer A for 5 column volumes or until UV absorbance reached baseline. Tau was eluted using a linear gradient (0-100%) of Buffer B (20 mM PIPES pH 6.5 with 1 M NaCl) over 20 column volumes and fractionated into 2 ml aliquots. Fractions were selected based on UV absorbance, resolved by SDS-PAGE and protein content was analyzed by InstantBlue Coomasie blue protein stain (Abcam). Those containing tau were collected, pooled and dialyzed (1:100) against Tau -SAA Buffer (10 mM Hepes pH 7.4 with 100 mM NaCl) in 10 K MWCO SnakeSkin Dialysis Tubing (Thermo Scientific) with constant stirring at 4 °C for four hours. A second round of dialysis was performed overnight. Following dialysis, the protein was centrifuged 15,000 x g at 4 °C for 15 min to remove any precipitate. The supernatant was then loaded onto pre-wetted 100 kDa MWCO Amicon Ultra-15 centrifugal filter units (Millipore Sigma) and centrifuged at 3,000 x g for 30 min at 4 °C. The retentate was discarded and the permeate pooled, aliquoted, and flash frozen in liquid nitrogen. The purity of the final product was accessed by Coomasie blue staining and the concentration determined by A280 on a Nanodrop One (Thermo Scientific) microvolume spectrophotometer. Additionally, final products were boiled with SDS sample buffer (Invitrogen), separated by SDS-PAGE, and immunoblotted with total tau antibody (TAU-5) (Thermo Fisher Scientific), anti-tau RD3 (EMD Millipore), and anti-tau RD4 (Cosmo Bio USA). Protein aliquots were stored at -80 °C until use.

Brain tissue samples and compliance with ethical standards

De-identified brain samples were obtained from the NIH Neurobiobank and its constituent repositories. Additional samples were obtained from Bispebjerg Hjernebank, Center for Neuroscience and Sterology, Bispebjerg Hospital as well as the UTHealth Brain Biorepository. Clinical diagnoses were confirmed by neuropathological examination. All samples were acquired from deceased, de-identified individuals with written informed consent for brain autopsy and use of materials for research purposes, and no additional ethical approval was required.

Tissue homogenization

Frozen brain tissue was removed from − 80 °C and placed directly onto dry ice. Samplse to be homogenized were dissected with a blade from a larger portion with a surgical feather blade with care to avoid white matter. After recording the sample mass (typically 100 mg were processed at a time) the tissue was minced with a blade and placed in a 5 ml dounce homogenizer on ice. The tissue was homogenized in 500 µl of phosphate buffered saline (PBS, Cytiva) with complete protease inhibitor (PI, Roche). Typically, ~ 10 strokes with the loose pestle, followed by ~ 10 strokes with the tight pestle were sufficient. The sample was transferred to a 1.5 or 2 ml micro tube and the volume was adjusted with PBS-PI to 10% (w/v). The sample was then intermittently probe sonicated using a vibra cell ultrasonic processor (Sonics) for 0.5–1 min at a power of ~ 2 W. The homogenates were aliquoted, flash frozen in liquid nitrogen, and stored in -80 °C.

Extraction of Tau filaments from AD brain

Sarkosyl-insoluble tau filaments were extracted as described by Fitzpatrick and colleagues [55] with minor modifications. Briefly, fresh frozen frontal cortex grey matter was homogenized in 20 volumes (w/v) of extraction buffer consisting of 10 mM Tris-HCl pH 7.5, 800 mM NaCl, 10% sucrose, 1 mM EGTA. Sarkosyl was added to 2% final concentration, after which the samples were incubated for 30 min at 37 °C. After a 20,000 x g centrifugation, the supernatant was collected and centrifuged again at 100,000 x g for 1 h at 4 °C. The resultant pellet was resuspended in extraction buffer (500 µl/g brain tissue) and centrifuged at 3,000 x g for 5 min. The supernatants were diluted three-fold in 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 10% sucrose and 0.2% sarkosyl and centrifuged at 166,000 x g for 30 min at 4 °C. The final sarkosyl-insoluble pellet was resuspended in 20 mM Tris-HCl pH 7.4 at 100 µl/g brain tissue.

Tau seed amplification assay (Tau-SAA)

Tau substrate was removed from − 80 °C and allowed to thaw on ice. After thawing, the protein was pooled and filtered once again using Amicon Ultra 0.5 ml centrifugal filters (Millipore Sigma). The protein concentration was re-measured using a Nanodrop One (Thermo Scientific) microvolume spectrophotometer. The tau protein was then added to a master mix consisting of Hepes 10 mM pH 7.4, 10 µM Thioflavin T (ThT), 150 mM NaCl, 120 µM Heparin (final concentration). The final concentration of tau, regardless of isoform utilized was 20 µM. To prepare the SAA reaction, 90 µl of the master mix was dispensed into wells of a black, clear-bottom 96-plate (Corning). Human brain homogenates used as seeds were allowed to thaw on ice, and were then sonicated for 2 rounds of 30 s (1 s on, 1 s off, amplitude:30) in an ice bath connected to an ultrasonic liquid processor (Misonix). Homogenates were then diluted in Tau Buffer (10 mM Hepes pH 7.4 with 100 mM NaCl) and 10 µl were spiked into designated wells of the SAA plate, giving a final volume of 100 µl in each well. After mixing, plates were sealed with Masterclear real-time PCR film (Eppendorf) and secured upon a Thermomixer (Eppendorf) for successive cycles of shaking and incubation. The thermomixer was placed inside a 37 °C incubator and set to 1 min of shaking at 800 rpm every 29 min. Fluorescence readings were taken using a SpectraMAX Gemini EM (Molecular Devices) spectrofluorometer at the start of each reaction (time: 0 h) and periodically throughout the length of the reaction (~ 200–300 h). Plates were read from the bottom using an excitation wavelength of 435 nm, emission at 485 nm and a gain of 350 V.

Pronase E digestion of Tau filaments

For the digestion of Tau-SAA amplified fibrils, reaction end products were harvested upon SAA completion. The contents of each well were agitated and collected into individual centrifuge tubes. Each sample was centrifuged at 20,000 x g for 30 min and the pellets were concentrated to 5 times the initial volume in PBS. Pronase E was added to 0.0001–0.1 mg·ml^− 1 final concentration. The reaction mix was incubated with constant agitation at 37 °C for 1 h. The sample was then diluted with SDS sample buffer and assessed by immunoblotting. For digestion of brain extracted filaments Benzonase (10 U/ml), MgCl₂ (2 mM) and DTT (50 mM) were added to the reaction mixture prior to pronase digestion.

Liquid Chromatography-Tandem mass spectrometry (LC-MS/MS) analysis of Tau-SAA-amplified fibrils

Pronase-digested Tau-SAA amplified fibrils were separated by SDS-PAGE and protein bands were excised for in-gel digestion following standard protocols [56]. An aliquot of the tryptic digest (in 2% actonitrile/0.1% formic acid in water) was analyzed by LC/MS/MS on an Orbitrap Fusion™ Tribrid™ mass spectrometer (Thermo Scientific™) interfaced with a Dionex UltiMate 3000 Binary RSLCnano System. Peptides were separated onto an analytical C18 column (100 μm ID x 25 cm, 5 μm, 18Å Reprosil-Pur C18-AQ beads from Dr Maisch, Ammerbuch-Entringen, Germany) at a flow rate of 350 nl/min. Gradient conditions were: 3%-22% B for 90 min; 22%-35% B for 10 min; 35%-90% B for 10 min; 90% B held for 10 min, (solvent A, 0.1% fomic acid in water; solvent B, 0.1% formic acid in acetonitrile). The peptides were analyzed using data-dependent acquisition method, Orbitrap Fusion was operated with measurement of FTMS1 at resolutions 120,000 FWHM, scan range 350–1500 m/z, AGC target 2E5, and maximum injection time of 50 ms; During a maximum 3 s cycle time, the ITMS2 spectra were collected at rapid scan rate mode, with HCD NCE 34%, 1.6 m/z isolation window, AGC target 1E4, maximum injection time of 35 ms, and dynamic exclusion was employed for 20 s.

Data processing and peptide mapping

Raw MS data files were processed using Thermo Scientific™ Proteome Discoverer™ (v1.4) and searched against the Uniprot-Human database using Sequest. Search parameters included a precursor mass tolerance of 10 ppm and a fragment mass tolerance of 0.6 Da. Carbamidomethylation of cysteine residues was set as a fixed modification, while oxidation of methionine, N-terminal acetylation, and phosphorylation of serine and threonine were included as variable modifications. No enzyme specificity was applied for peptide mapping. False discovery rates (FDR) were set to 1% for strict identification and 5% for relaxed identification using Percolator.

Identified peptides were mapped to the 2N4R tau isoform (Uniprot P10639-8), and the start and end positions of each peptide were determined. Peptide-spectrum matches (PSM) were filtered at a q-value threshold of ≤ 0.01 to retain high-confidence identifications. To quantify residue-level abundance, PSM values were normalized within each sample by dividing by the total PSM detected in that sample. Normalized PSM values were then summed across residues to determine residue-level abundance. The mean abundance per residue was calculated by averaging normalized PSM values across three independent AD brain-derived SAA fibril samples.

Tau-SAA for drug screening

Suspected aggregation inhibitors were acquired from several sources, including Sigma Aldrich, Santa Cruz Biotechnology, ApexBio, GFS Chemicals, Chem-Impex, and the Cayman Chemical Company. For a detailed list of chemical, sources please refer to supplementary Table (11) Compounds were purchased dissolved in dimethyl sulfoxide (DMSO) or as powders that were then dissolved in DMSO (Sigma-Aldrich). CNS-penetrant compounds were purchased from Selleck chemical filtering FDA-approved compounds for brain permeability. Please refer to supplementary file (2) All compounds were dissolved in 100% DMSO at a 10 mM final concentration. Stocks were diluted to 1 mM in nano-pure water and 10 µl were added to the Tau-SAA reaction mix (100 µM final concentration). As a vehicle, 10% DMSO in water was added. All compounds were tested in triplicate. The area under the curve (AUC) from the mean of three replicates was calculated using built-in analysis functions in Graph Pad Prism 9. Data was displayed as a percentage of vehicle AUC. To test various concentrations, compounds were further diluted in 10% DMSO. Each drug concentration was tested in triplicate and data plotted as mean ± SEM. A line of best fit was determined using the built-in Boltzman-Sigmoidal non-linear regression function in Graph Pad Prism. To determine the half-maximal inhibitory concentration (IC₅₀), the AUC was determined for each drug concentration and expressed as a percentage of the vehicle AUC. IC₅₀ was determined using the log(inhibitor) vs. response -- variable slope analysis function in Graph Pad Prism.

Sedimentation assay

Tau SAA products were harvested at the completion of the SAA reaction. Three replicates were pooled and 100 µl were subject to centrifugation at 100,000 x g for 1 h. The supernatant was aspirated without disturbing the pellet, and the pellet was resuspended in an equal volume of phosphate-buffered saline (PBS). The total, supernatant, and pellet fractions were boiled in SDS-PAGE sample buffer and separated by SDS-PAGE, followed by immunoblotting with tau monoclonal antibody (TAU-5) (ThermoFisher Scientific).

Transmission electron microscopy

Samples were briefly sonicated in a water bath sonicator (Misonix Model S-4000) at an amplitude of 20 for 20 s. Formvar/Carbon 200 Mesh Copper grids (Electron Microscopy Sciences) were briefly rinsed with a drop of ddH2O and then incubated with the sample for 60 s. The grids were then rinsed with another drop of ddH2O and stained with three sequential drops of 4% uranyl acetate (Electron Microscopy Sciences) for a total of 60 s. Following staining, grids were rinsed twice with drops of ddH2O and then allowed to air-dry under a cover disk for 10 min. All drops were 5 µL and blotted with Whatman filter paper. Grids were imaged using a JEOL JEM-1400 transmission electron microscope at 120 kV.

Immunoblotting

Western blotting was carried out using a standard protocol as previously described [57]. Samples were resolved on 4–12% or 12% Bis-Tris gels (Invitrogen) and nonspecific binding blocked with 2% nonfat milk in PBS with 0.1% Tween 20 (PBS-T). For assessing purified isoform constructs, antibodies TAU-5 (Thermo Fisher Scientific, #AHB0042), anti-tau RD3 (EMD Millipore, #05-803), and anti-tau RD4 (Cosmo Bio USA, #CAC-TIP-4RT-P01) were used at 1:4000 dilution. For epitope mapping, antibodies HT7 (Thermo Fisher Scientific, #MN1000), TAU-5, 77G7 (Biolegend, #816701), C4 (Sigma Aldrich, #ABN2178), RD3, A16097D (Biolegend, 851001) and Tau-46 (Cell Signaling Technology, #4019) were used at a 1:2000. All antibody epitopes were recovered from the manufacturer’s website.

Results

Designing an ultrasensitive and specific Tau-SAA

In the human central nervous system (CNS), tau manifests as six isoforms resulting from the alternative mRNA splicing of the MAPT gene [58]. Our initial efforts concentrated on investigating and identifying the isoform most suitable as a substrate in the SAA format. To this end, we generated full-length, cysteine-free constructs without any foreign tags, expressed them in E. coli, and purified the resulting tau proteins using a simple one-step cation-exchange chromatography protocol. Purification products exhibited similar purity as assessed by Coomassie Blue staining (Figure S1A), irrespective of the specific tau isoform. All isoforms gave positive signals at expected molecular weights when immunoblotted with total tau antibody (Figure S1B, top panel), while only 3R isoforms were visualized upon probing with a 3R isoform-specific antibody (Figure S1B, middle). Likewise, a 4R-specific antibody was used to visualize the three 4R isoforms (Figure S1B, bottom).

To assess the suitability of each construct as a substrate for SAA, we conducted ten-fold serial dilutions of AD or healthy control (HC) brain homogenate (BH) from 10^− 4 (10,000-fold) to 10^− 9 (1 billion-fold). An equimolar concentration of each tau isoform was supplied as a substrate and the resulting SAA reaction was monitored for ThT fluorescence over 300 h. Interestingly, while none of the isoforms displayed significant aggregation when treated with HC BH, 3R tau isoforms exhibited a greater capacity than their 4R counterparts in amplifying AD brain-derived tau seeds (Fig. 1). Amongst the 3R isoforms, the ability to detect progressively higher dilutions of AD brain appeared to diminish with the inclusion of N-terminal domains (Fig. 1, top panel). Correspondingly, the maximum ThT value and signal-to-noise ratio in AD brain-seeded reactions appeared to be inversely related to the presence of N-terminal domains (Fig. 1, top panel). Using 0N3R as a substrate, we successfully detected complete AD brain homogenate seeding activity down to a dilution of 10^− 8 in all three replicates and partial activity (1/3 replicates) at 10^− 9. The lag time, representing the duration before the exponential phase of the aggregation curve, increased with successively higher dilutions, illustrating a quantitative relationship between the amount of seeds in the AD BH inoculum and the onset of seeding activity. Altogether, we concluded that the 0N3R isoform exhibits the greatest aptitude as a substrate for Tau-SAA.

Fig. 1.

Tau isoform composition influences Tau-SAA efficiency. The six human CNS tau isoforms were independently assessed as substrates for Tau-SAA. Brain homogenate (BH) from one AD and one control case were serially diluted (from 10^− 4 to 10^− 9) and used to spike the Tau-SAA reaction mix. The ensuing reaction was monitored by ThT fluorescence for ~ 300 h. Each dilution was tested in triplicate and data displayed as the mean ± SEM

As our initial experiment utilized only a single AD and HC sample, we next sought to evaluate a larger cohort of AD and control patients. We acquired brain samples from the prefrontal cortex (BA46) of 11 AD patients (Braak III-VI) and 15 controls (Braak 0-III) (Table S1, detailed sample information can be found in Supplementary File 1). Serial dilutions of brain homogenates (from 10^− 5 to 10^− 9) were tested in triplicate for each of the 26 samples. To set a stringent threshold, we considered a replicate positive only if it surpassed a florescence value of 1000 (au) over 200 h of run time. Samples were deemed positive at a given dilution only if two or three of the three technical replicates surpassed the 1000 au threshold. As before, seeding activity appeared related to the dilution of AD brain homogenate used to seed the reaction in a dose-dependent manner. At a dilution of 10^− 5, 10 out of 11 (91%) AD BHs were deemed positive (Fig. 2). At higher dilutions of 10^− 8 or 10^− 9 the number of positive samples was diminished to 5 and 4 out of 11 (45% and 36%, respectively) (Fig. 2). Of the 15 healthy control brain homogenates assessed, only one (7%) sample was deemed positive at 10^− 5, the lowest dilution tested (Fig. 2, Supplementary File 1). Based on these results, we concluded that Tau-SAA can effectively discriminate between AD and control BH.

Fig. 2.

High sensitivity of Tau-SAA and high specificity for AD over controls. Postmortem brain samples (prefrontal cortex; BA46) from 26 different patients were tested with Tau-SAA using serial dilutions from 10^− 5 to 10^− 9. Graphs display the median of three technical replicates for each patient. Samples were deemed positive only if at least two out of three replicates exceeded a ThT fluorescence value of 1000au. The number of samples deemed positive at each dilution is indicated in the top left corner of each graph

To further assess assay specificity and explore the possibility of cross-seeding, we challenged the Tau-SAA with recombinant α-synuclein preformed fibrils (αSyn PFFs). While recombinant tau PFFs readily seeded Tau-SAA reactions at all tested concentrations (200-0.2ng), αSyn PFFs failed to induce tau seeding in our assay, even when supplied at high concentrations (200ng) and resembled unseeded buffer-alone reactions (Figure S2A). Moreover, we tested frontal cortex BHs from 9 neuropathologically diagnosed Parkinson’s disease (PD) and 10 multiple system atrophy (MSA) patients across the same range of dilutions. None of the MSA samples displayed seeding activity and only one PD sample (11%) tested positive at the lowest dilution tested (10^− 5). Taken together, these findings support the specificity of the Tau-SAA for tau aggregates.

Tau-SAA fibrils exhibit a similar core composition to AD-derived Tau filaments

Proteolytic resistance is a widely used biochemical approach for characterizing conformational strains. When exposed to proteases, different aggregate conformations yield distinct fragmentation patterns, which can be visualized through immunoblotting [57, 59, 60, 61]. These differences are thought to arise from variations in the three-dimensional structure that either permit or restrict proteolytic cleavage at specific sites.

To examine the proteolytic profile of Tau-SAA fibrils, we used three distinct AD brain samples to seed Tau-SAA and collected the resulting fibrils at the end of the reaction. For comparison, we extracted tau filaments from the corresponding brain samples. Both brain-derived and SAA-amplified fibrils were digested with Pronase E, denatured, and immunoblotted using an anti-Tau RD3 antibody. As expected, both brain-derived and amplified tau fibrils were resistant to proteolysis, producing similar protease-resistant fragments, with the smallest one migrating between 6 and 14 kDa (Fig. 3A). This pattern was consistently observed across all three brain-derived samples and their SAA-amplified counterparts (Fig. 3A) and aligns with previous studies of the AD tau protease-resistant core [55]. Interestingly, brain-derived tau samples also displayed a prominent band between 17 and 28 kDa, and in two of three samples, higher molecular weight smears were present, potentially indicative of dimers, multimers, and other higher-order species resistant to denaturation or incompletely digested tau fragments originating from other endogenous tau isoforms (Fig. 3A).

Fig. 3.

Protease resistance and mass spectrometry of AD brain-amplified tau fibrils. (A) Three AD cases (AD1-3) were used to seed Tau-SAA. The amplified fibrils (SAA) as well as fibrils extracted directly from brain tissue (Brain) were treated with Pronase E (100 µg/ml) and immunoblotted with anti-Tau RD3 antibody (Sigma Aldrich 05-803) which is reactive against residues 267–316 (R1 and R3) omitting the R2 repeat domain where it bridges R1 and R3. (B) Schematic representation of key regions of the 0N3R tau isoform (Uniprot P10636) utilizing the 2N4R (P10636-8) numbering. Proline-rich regions 1 and 2 are indicated by P1 and P2. Repeat domains R1, R3, and R4 are labeled as such. Black lines represent regions that are omitted in the 0N3R isoform. The amino acid sequence is continuous despite gaps in the numbering and diagram. (C) Peptide abundance and amino acid coverage of the pronase-resistant core of SAA-amplified filaments from three different brain samples (AD1–3). Amino acid position is indicated on the bottom x-axis, whereas the y-axis represents the abundance expressed as a percentage of total peptide spectrum matches (PSMs) for that sample. Dashed grey lines indicate domain organization and align schematic shown above (B). The secondary (top) x-axis marks the start and end positions of key regions in brain-derived tau filaments, including the residues that comprise the AD tau fold protofilament core (306–378, blue) and the pronase-resistant core of brain-derived 3R tau (260–406, red) identified by Fitzpatrick and colleagues [55]

Cryo-EM studies of AD brain-derived tau have mapped the protofilament core to the R3 and R4 repeat domains (residues 306–378) [55]. The same landmark study also identified a less-ordered β-sheet region that could extend 16 additional residues, corresponding to residues 259–274 (R1) from 3R tau or 290–305 (R2) from 4R tau, that remained after pronase digestion [55]. To further characterize the amino acid composition of our SAA-amplified fibrils, we performed epitope mapping using a panel of antibodies targeting distinct tau regions. To determine whether N-terminal regions of the protein remained in the SAA-amplified fibril core after pronase digestion, we utilized anti-tau HT7 (Invitrogen) and Tau 5 (Invitrogen) monoclonal antibodies, which recognize the proline-rich P1 and P2 regions. While both antibodies recognized undigested SAA-amplified filaments, neither recognized the protease-resistant core after digestion with 0.1 mg∙ml^− 1 Pronase E, indicating that these regions are absent in the SAA fibril core (Figure S3A). Similarly, C-terminal antibodies, A16097D (Biolegend) and Tau 46 (Cell Signaling Technology) failed to mark the digested core, suggesting the exclusion of the C-terminal region (Figure S3A). To target the R3 and R4 domains, previously identified in AD brain-derived filaments [55], we used C4 (EMD Millipore) and 77G7 (Biolegend) antibodies. Both labeled a fragment between the 6 and 14 kDa markers after digestion of SAA-amplified fibrils. Collectively, our antibody panel maps the SAA-amplified pronase-resistant fragment between residues 267 to 369, omitting the R2 repeat domain absent in our 0N3R SAA substrate (Fig. 3B, Figure S3A).

To more precisely characterize the amino acid composition of the SAA-amplified filament core, we performed complementary mass spectrometry analysis of pronase-treated SAA fibrils amplified from three distinct AD brains (Fig. 3C, Figure S3B). The pronase-resistant fragments were excised from gels, subjected to trypsin digestion, and analyzed by liquid-chromatography and mass spectrometry (LC-MS). Peptides were mapped to the 0N3R tau isoform using full-length CNS tau (2N4R) numbering (Fig. 3B). Mass spectrometry analysis revealed that the SAA pronase-resistant core spans residues I₂₆₀ – R₄₀₆, precisely matching the composition reported for brain-derived filaments [55]. The greatest proportion of peptides mapped to the R3 and R4 domains, consistent with the R3:R4 protofilament core, with a lower abundance of peptides extending N-terminally to R1 and C-terminally to R₄₀₆ (Fig. 3B-C, S3B), in agreement with the presence of a less-ordered β-sheet region. Notably, the most abundant identified peptide, IGSLDNITHVPGGNK (359–369) (Figure S3B), mirrored the most abundant residues found in brain-derived filaments [55]. Importantly, no residues mapped to regions before 260 (R1) or beyond 406, or to the R2 domain, absent in our SAA substrate. Although peptide abundance varied, all residues from 260 to 406 were detected with the exception of a seven-residue stretch in R3 from 318 to 324 (VTSKCGS), presumably a consequence of the cysteine-to-serine point mutation in our construct (Figure S3B).

Application of Tau-SAA to screen for aggregation inhibitors

The spread of pathological tau species through neuroanatomically interconnected and communicating neurons is thought to follow a prion-like seeding mechanism wherein misfolded tau species induce the misfolding of natively folded proteins [21, 27]. The ability of Tau SAA to recapitulate this conversion in an in vitro setting presents a unique opportunity to screen for therapeutic compounds and small molecules capable of neutralizing or diminishing the seeding potential of AD brain-derived aggregates. We hypothesized that the Tau-SAA platform could be adapted into an effective, scalable, and high-throughput drug screening tool.

To investigate this potential, we conducted a literature-based selection of 21 compounds previously reported as inhibitors of amyloid and/or tau aggregation (Table S2), hereafter referred to as suspected aggregation inhibitors (SAI). This group included compounds such as rhodanine [63], the phenothiazines azure A, thionin, and quinacrine [64], as well as the naphthelenesulfonate dye Congo red [65]. To complement these compounds, we hand-selected an additional 201 bioactive FDA-approved drugs with demonstrated CNS penetrance (Selleck Chemicals) (Supplementary File 2). This compound library, hereafter referred to as CNS-p, has undergone extensive clinical evaluation for a variety of clinical applications, including asthma, inflammation, type 2 diabetes, memory impairment, as well as the treatment of certain cancers. Grouping compounds by targeted signaling pathway revealed that neuronal signaling was the most common target (61/201, 30%), with other notable pathways including metabolism (10%), microorganism (10%), and immunology and inflammation (6%) (Figure S4A).

To screen these libraries for tau seeding inhibitors, we conducted Tau-SAA in the presence of each compound at a final concentration of 100µM. As before, ThT fluorescence served as a readout for tau aggregation and was monitored over the course of 300 h (Fig. 4A). As a preliminary screen, we assessed a subset of compounds (CNS-p: 29 and SAI: 5) for inhibitory activity using ten-fold serial dilutions of AD brain homogenate from 10^− 4 to 10^− 8 (Figure S5). We reasoned that the inhibitory effects of compounds on tau aggregation could manifest in two ways, (1) an increase in lag time or time to threshold (TTT), indicative of slower aggregation or (2) a decrease in maximum thioflavin fluorescence (ThT Max), potentially indicative of reduced fibril formation. To serve as a comparison, we included a vehicle control (1% DMSO) without any compound. Of the five SAI compounds assessed, three (Azure A, Thionin, and Congo Red) exhibited almost complete inhibition (TTT > 300 h; Figure S5A) irrespective of the dilution of AD brain homogenate used to seed the reaction. Correspondingly, these compounds had ThT Max values approaching the baseline or the value at reaction onset (Figure S5B). An additional SAI compound, quinacrine hydrochloride, appeared to only delay TTT at higher dilutions of AD brain homogenate (10^− 6-10^− 8) but reduced ThT Max across all dilutions (Figure S5A-B). Of the 29 CNS-p compounds, only the antimicrobial drug Rifampin and breast cancer chemotherapy agent Epirubicin exhibited a similar complete inhibitory effect in terms of TTT as well as ThT Max (Figure S5).

Fig. 4.

AD Brain Templated Tau-SAA efficiently identifies tau aggregation inhibitors. (A) Schematic illustration of the application of Tau-SAA as a drug screening platform. (B) Each compound was tested in triplicate at a 100 µM final concentration. AD brain homogenate (10^− 5 final dilution) was used to seed the reaction. The area under the curve (AUC) was calculated using the raw ThT traces for each compound and expressed as a percentage of the vehicle (1% DMSO) AUC. A total of 222 compounds were assessed (mean AUC 83.4 ± 2.7%). (C) Compounds from each library were categorized into four groups based on percent inhibition (< 0, 0–50, 50–75, 75–100%). (D) Compounds with percent inhibition > 75% from both SAI and CNS-p libraries. Selected compounds are labeled. SAI: suspected aggregation inhibitors. CNS-p: CNS-penetrant and FDA-approved compounds from Selleck Chemicals

Given that limited differences were observed across distinct dilutions of AD brain homogenate, we selected to assess the remainder of both the SAI and CNS-p libraries at a single dilution of seed (10^− 5). Each compound was tested at 100µM final concentration. To establish a quantifiable metric that integrates both TTT and ThT Max, we calculated the area under the curve (AUC) for raw ThT traces (mean of three technical replicates) and expressed the data as a percentage of vehicle AUC (Fig. 4B, Table S2, Supplementary File 2). Collectively, the 222 compounds exhibited a mean AUC of 83.4 ± 2.7% of the vehicle (Fig. 4B).

To further interpret the results, we calculated the percent inhibition for each compound by subtracting the percent AUC for each compound from the AUC of the Vehicle (100%), and binned the compounds into four groups, those with percent inhibition less than 0 (i.e. promoters of tau aggregation), (AUC > 100% of vehicle), those with 0–50% inhibition, 50–75%, and 75–100% inhibition (Fig. 4C). A total of 59 compounds from the CNS-p library as well as two SAI exhibited an AUC greater than 100% of the vehicle or a negative percent inhibition (Fig. 4C), indicating that they augmented tau aggregation either by decreasing TTT, increasing ThT Max, or some combination of the two parameters. The majority of CNS-p library compounds (119/201, 59%), as well as five SAI compounds showed minor inhibition (0–50%) relative to the vehicle. Additionally, 8% of CNS-p compounds (16/201) showed 50–75% inhibition as well as two SAI compounds. Collectively, 19 compounds, including seven (7/201, ~ 3%) from the CNS-p library, exhibited inhibition greater than 75% (Fig. 4C-D). These included, (i) tolcapone (78% inhibition), a selective reversible catechol-type inhibitor used to treat Parkinson’s disease, (ii) (+)-catechin (89%), a naturally occurring flavonol, (iii) topotecan (92%), an antineoplastic chemotherapy drug, (iv) honokiol (94%), a polyphenol, (v) sunitinib (98.6%), a targeted cancer drug as well as (vi) rifampin (> 99%) and (vii) epirubicin (> 99%).

Expectedly, SAI compounds were more effective at inhibiting AD brain-derived seeding activity, with more than half (12/21, 57%) of examined compounds inhibiting over 75% of seeding activity (AUC ≤ 25% of vehicle) (Fig. 4C-D). Across both libraries, seven compounds (7/222, 3.6%) inhibited over 99% of tau aggregation activity (AUC ≤ 1% of vehicle). Phenothiazines from the SAI library made up the largest constituent of this subset (thionin, methylene blue, azure a-c).

Validation of Tau-SAA drug screening

As a secondary validation, we selected three drugs based on a combination of their inhibitory activity in our assay and literature support for their anti-aggregation potential. Indomethacin, a nonsteroidal anti-inflammatory drug, was selected as a negative control due to its limited inhibition of tau aggregation in our study (17.3 ± 9.3%) and lack of literature support. Rifampin (99.9 ± 5.1%), an antimicrobial agent, served as a positive control, with prior reports supporting its anti- tau aggregation effects [66, 67]. Sunitinib, a kinase inhibitor employed for the treatment of advanced or metastatic kidney cancer, was selected as a strong inhibitor identified in our assay, (98.6 ± 7.9%) with no literature evidence of tau or amyloid aggregation inhibition.

Tau-SAA products from these reactions, as well as the untreated vehicle (Fig. 5A), were extracted from the 96 well plates and subjected to sedimentation at 100,000 x g. The total fraction (T) before centrifugation, supernatant (S), and pellet (P) fractions were separated by SDS-PAGE and immunoblotted with total tau antibody (Fig. 5B). In the untreated vehicle condition, sedimentation revealed the presence of insoluble tau species indicative of large, aggregated fibrils (Fig. 5B). Similar results were obtained with indomethacin, in concordance with the limited inhibition observed in Tau-SAA (Fig. 5A-B). Conversely, SAA products from rifampin treated reactions remained in the soluble fraction, suggesting that the compound inhibit the formation of large aggregates. Interestingly, the SAA product from sunitinib-treated wells contained insoluble tau species that appeared in the pellet fraction (Fig. 5B, right) despite an apparent lack of ThT fluorescence during the Tau-SAA (Fig. 5A, right).

Fig. 5.

Validation of Tau-SAA drug screening. (A) Raw ThT traces for vehicle (1% DSMO), indomethacin, rifampin, and sunitinib screening at 100 µM. Three individual replicates are displayed. (B) Sedimentation assay of Tau-SAA drug screening end-products. SAA products were harvested and centrifuged at 100,000 x g for 1 h. The total, supernatant, and pellet fractions were separated and immunoblotted with total tau antibody (TAU-5). (C) Transmission electron microscopy (TEM) micrographs of Tau-SAA end products (total fraction). Low magnification (top panel) images, scale bar 1 μm. The bottom panel displays high magnification of the same sample (scale bar 200 nm)

To further confirm the presence of aggregates, Tau-SAA products were also examined by transmission electron microscopy (Fig. 5C). Consistent with the results of the sedimentation assay, widespread amyloid-like aggregates (Fig. 5C, top) were observed in SAA products from vehicle, indomethacin, as well as sunitinib reactions (Fig. 5C). Fibrils appeared to be composed of two protofilaments with regular twisting (Fig. 5C, bottom). Rifampin Tau-SAA product was devoid of fibrillar amyloid structures.

Tau-SAA can be used to estimate half-maximal inhibitory concentration (IC₅₀)

Our initial screening of both SAI and CNS-p libraries identified 19 compounds that inhibited over 75% of seeding activity from AD brain-derived tau seeds when added to Tau-SAA at 100µM concentration (Fig. 4C-D).

To explore the potency of these compounds, we selected two compounds from the SAI library (Methylene Blue, and Exifone), as well as two from the CNS-p library (Rifampin and Epirubicin) for a comprehensive dose-response analysis. Dilutions spanning from the initial 100µM concentration down to 0.1µM final were prepared and introduced at the onset of the SAA reaction. The inhibitory effects on tau seeding activity were monitored over time using ThT fluorescence as a readout and plotted as a percentage of vehicle ThT Max (Fig. 6, top). For all four compounds, ThT fluorescence exhibited a clear decrease with increasing dosage. As a quantitative measure, half-maximal inhibitory concentration (IC₅₀) was derived by calculating the AUC of the raw ThT trace for each dosage (expressed as a percentage of the vehicle AUC) (Fig. 6, bottom). Among the compounds selected for dose-response analysis, Methylene Blue displayed the most robust inhibition (IC₅₀: 0.3µM), followed by Epirubicin (6.6µM), Exifone (11.2µM), and Rifampin (11.8µM) (Fig. 6B).

Fig. 6.

Dose-response analysis of tau aggregation inhibitors. Top: Tau-SAA fluorescence traces of several compounds, each tested at various concentrations (between 100 − 0.1 µM final) are displayed as a percentage of the maximum ThT fluorescence intensity reached by the Vehicle-treated condition (mean ± SEM). The line of best fit was determined using a Boltzmann-Sigmoidal non-linear regression. Each condition was tested in triplicate. Bottom: Half-maximal inhibitory concentration (IC₅₀) was determined using a log(inhibitor) vs. response, variable slope function (Graph Pad Prism v9) for each condition and expressed as a percentage of the area under the curve (AUC) in the vehicle-treated condition. IC₅₀ values are displayed in the top right corner of each graph

Discussion

The prion-like propagation of misfolded tau species is believed to drive the stereotypical progression of tau pathology throughout the brain and is a key pathognomonic feature of AD and other tauopathies [29, 58]. The development of therapeutic strategies targeting pathological tau remains a priority in the field [68, 69, 70, 71]. Seed Amplification Assays (also known as PMCA or RT-QuIC) emulate the prion-like conversion of natively folded protein in a cell-free in vitro platform, enabling the amplification of ‘seeds’ present in biological samples with exceptional sensitivity and specificity [47, 72, 73]. The ability to detect even minute quantities of misfolded aggregates has already highlighted the potential of SAA as a tool for early diagnosis of Parkinson’s disease and Creutzfeldt-Jakob disease [74, 75, 76].

In the present study, we developed a Tau-SAA and evaluated its performance using 26 postmortem brain samples from neuropathologically diagnosed AD and control subjects. Given recent reports suggesting that tau aggregation can be influenced by alpha-synuclein cross-seeding [77, 78, 79], we also examined the seeding capacity of 19 distinct brain samples from synucleopathies (Figure S2B). Our results showed that alpha-synuclein aggregates both prepared in vitro or derived from patients’ brains did not significantly induced Tau aggregation under our experimental conditions.

During the development of our Tau-SAA, we prioritized the use of physiologically relevant tau isoforms as substrates. Early iterations of Tau-SAA relied largely on artificial tau constructs or fragments as substrates [45, 46, 47, 48]. With this consideration, we established a simple and easily applicable one-step cation exchange chromatography to purify all six CNS tau isoforms, eliminating the need for foreign protein or poly-histidine tags. Testing each isoform revealed that 3R isoforms exhibit faster seeding kinetics and higher Thioflavin T signal-to-noise ratios than their 4R counterparts. These findings align with a recent study that similarly tested all six full-length isoforms and observed increased seeding activity in 3R isoforms compared to 4R [50]. The same study, as well as our own results, demonstrated decreased seeding activity upon the addition of a single N-terminal domain (1N3R) and an even greater detrimental effect upon the inclusion of both N-terminal domains (2N3R) [50]. Another independent study similarly concluded that the 0N3R isoform is the optimal substrate for amplification of tau seeds from AD brain [52].

While the tau isoform appears to play a critical role in amplification efficiency, multiple lines of evidence suggest that the choice of tau isoform may not be a critical determinant of whether fibrils can adopt disease-relevant conformations [55, 80, 81]. Recent solid-state NMR studies of AD brain-templated fibrils further demonstrated that isotopically labeled 4R and 3R isoforms readily intermix (60:40 ratio) within fibrils templated with AD brain-derived tau, indicating that both forms are readily recruited into the pathological fold of AD tau fibrils [82].

A major objective of this study was to evaluate whether Tau-SAA-amplified fibrils resemble AD brain-derived filaments. While several groups have demonstrated that SAA can propagate tau seeds from AD brain, the biochemical composition of these amplified fibrils has not yet been thoroughly characterized. To begin characterizing the biochemical properties of SAA-amplified fibrils, we examined their proteolytic resistance to gain insights about the identity of the fibrillar core. Pronase digestion revealed that SAA-amplified fibrils share a protease-resistant core similar in size to AD brain-derived aggregates. Complementary epitope mapping and mass spectrometry analysis further confirmed that the resistant core of SAA-amplified filaments exactly mirrors that of AD-derived tau filaments, supporting their structural relevance. Our studies report the first mass spectrometry-based characterization of SAA-amplified fibril protease-resistant cores, revealing that their amino acid composition is identical to that of brain-derived tau. Despite this clear biochemical similarity, we acknowledge that sharing the same fibrillar core does not necessarily imply identical structure. Additional high-resolution structural methods, such as cryo-EM, will be necessary to determine whether Tau-SAA amplified fibrils truly faithfully replicate the structural features of AD-derived tau filaments. Nonetheless, our findings provide strong preliminary evidence that Tau-SAA propagates a disease-relevant tau core, reinforcing its potential as a model for studying pathological tau seeding and aggregation. Consequently, the application of Tau-SAA as a drug screening platform is highly relevant towards the discovery of compounds that prevent tau fibril formation or propagation.

Leveraging the ability to amplify brain-derived tau aggregates, we employed the Tau-SAA as a platform for screening small-molecules for inhibition of tau seeding. In total, we screened 222 compounds from FDA-approved, blood-brain barrier permeable libraries, as well as hand-selected small molecules previously described as tau or amyloid inhibitors. To the best of our knowledge, this study represents the first application of SAA as a platform for the screening of tau aggregation inhibitors. Importantly, we utilized AD brain-derived tau aggregates as seeds for drug-screening Tau-SAA, diverging from the common practice of using spontaneously formed recombinant tau fibrils, which have been shown to adopt a different structure than brain aggregates [37].

To identify compounds capable of inhibiting AD brain-templated tau seeding, we screened two compound libraries: (i) 21 suspected aggregation inhibitors (SAI) identified through a literature review [63, 64, 65, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94], and (ii) 201 FDA-approved, CNS-penetrant compounds purchased from Selleck Chemicals (CNS-p) (Supplementary File 2). As expected, SAI compounds collectively exhibited strong inhibitory activity, with the majority reducing tau aggregation and 12 compounds inhibiting > 75% the seeding activity. Several exhibited near-complete inhibition (> 95%), including azure A, azure B, azure C, Congo red, myricetin, rosmarinic acid, methylene blue, exifone, and baicalein. While these strong inhibitory effects align with previous studies, other SAI compounds produced less pronounced inhibitory effects. Orange-G, a small molecule dye known to inhibit the aggregation of amyloid-β in vitro [87] only mildly reduced (39%) tau seeding. Similarly, usnic acid, a secondary metabolite in lichen [31], whose derivatives have been shown to inhibit self-fibrillization of a tau hexapeptide and spontaneously generated tau 2N4R fibrils [88] proved largely ineffective against AD brain-templated tau aggregation (21.5% inhibition). Likewise, while fulvic acid has been reported to inhibit spontaneous tau fibril formation [90], it exhibited only a modest inhibitory effect (33.4%) in our assay. Interestingly, anle138b was the least effective SAI compound tested (-25.9% inhibition), and instead enhanced tau seeding activity. Anle138b is a brain-penetrant diphenyl pyrazole that has shown efficacy as an inhibitor of α-synuclein and prion protein oligomerization and proved effective in mouse models [95]. Further studies have shown that anle138b also inhibits the in vitro formation of spontaneous 2N4R aggregates, reduces hyperphosphorylated tau burden and prolongs survival in the PS19 tau mouse model [94]. This discrepancy underscores that while immensely useful as screening tools, in vitro cell-free platforms may not replicate all aspects of drug-tau interactions.

Compared to SAI compounds, Tau-SAA screening of compounds from the CNS-p library yielded significantly fewer inhibitory compounds. Among the few compounds that inhibited more than 75% of tau seeding activity, many had supporting evidence in the literature. Honokiol, a polyphenol used in traditional Chinese herbal medicine [96], likely derives its anti-aggregation properties from its biphenyl backbone. While its direct effects on tau aggregation have not been reported, its anti-amyloidogenic activity is well documented [96]. Topotecan, a potent topoisomerase 1 inhibitor approved for the treatment of various types of cancer, was previously reported to suppress mutant huntingtin aggregation in a mouse model, although the authors concluded that this was likely caused by downregulated huntingtin gene expression [97]. Another study reported that topotecan only mildly inhibited the fibrillization of tau peptides [98]. The inhibitory effects of catechins, polyphenolic bioactive compounds naturally occurring in green tea, have also been well described [99]. Among the most potent inhibitors identified in our screen, rifampin, has been shown to limit the spread of tau oligomers and improve cognitive function in mouse models [66, 67, 100]. Epirubicin, a catechol-containing antineoplastic compound, has also been recently described as an inhibitor of protein aggregation [101]. The identification of these compounds among 222 screened drugs highlights the utility of Tau-SAA as a useful and widely applicable drug screening tool.

Sunitinib also displayed a seemingly high level of inhibition in Tau-SAA, yet secondary validation demonstrated the presence of insoluble tau species as well as tau filaments. We anticipated two potential scenarios in which Tau-SAA could mistakenly indicate a lack of tau aggregation, resulting in a false positive result: (1) drug interference with ThT binding, (2) interference with ThT excitation or emission. This false positive result underscores the need for secondary validation of ‘hits’ from Tau-SAA or any drug screening platform.

Our findings suggest that Tau-SAA can serve as a robust screening method for identifying small molecules that inhibit or entirely prevent the aggregation of tau when exposed to a biologically relevant source of tau seeds. Through thoughtful modifications, the assay can be converted into a high-throughput screening tool, enabling the rapid screening of exceedingly large compound libraries. Furthermore, through screening of FDA-approved compounds, Tau SAA offers a valuable strategy for drug repurposing, potentially accelerating the discovery of clinically viable tau-targeting therapies.

Limitations

The use of Tau-SAA for drug screening has the potential limitation that the structure of the tau aggregates produced in the assay might be different from the aggregates present in AD brain. Our data with epitope mapping and mass spectrometry of the core filaments generated by Tau-SAA suggest that the main biochemical properties are kept during amplification. However, we acknowledge that these biochemical studies do not necessarily imply an identical structure of tau aggregates. Additional high-resolution structural studies, such as cryo-EM, will be necessary to determine whether Tau-SAA amplified fibrils truly replicate the structure of AD-derived tau filaments. Another limitation is the relatively small sample size of AD and control cases used in this study, which limits our ability to conclude that Tau-SAA can detect tau aggregates in all AD cases. Furthermore, it would be necessary to study other tauopathies to determine whether the Tau-SAA is specific for AD or can detect tau aggregates in other tau-related diseases.

Conclusions

Our findings indicate that Tau-SAA can effectively detect tau aggregates from AD up to a dilution of 100 million-fold of the brain homogenate. Moreover, the assay can be adapted to screen compounds capable of inhibiting the seeding and aggregation of tau. Since tau spreading by seeding is one of the hallmark events in AD pathogenesis, these compounds might be useful as candidates for therapeutic intervention.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

D.G. performed the majority of the experiments, analyzed the data, drafted the figures and wrote the first draft of the manuscript. H.E. performed experiments and assisted in data interpretation and preparation of figures. T.A., D.H., and V.B. conducted transmission electron microscopy experiments. C.B. assisted in protein purification. N.M aided in the design of the Tau-SAA. S.K., J.F., S.A. and P.S. aided in sample acquisition and interpretation of results. F.W. and M.S. supervised the research and interpretation of results. M.S. and C.S. provided funding for the study. C.S. conceived the original idea and supervised the design and implementation of the research. All authors reviewed and contributed to the final version of the work.

Funding

The study was supported by NIH grants R61 NS136666, R01AG059321 and R01 AG055053 to CS and a grant from the Texas Alzheimer’s Research Care and Consortium (TARCC) to MS.

Data availability

All Tau-SAA drug screening data generated in this study are publicly available in the PubChem database (BioAssay AID: 2061166). Any request for materials should be made to the corresponding author.

Declarations

Ethics approval and consent to participate

All samples were acquired from deceased, de-identified individuals with written informed consent for brain autopsy and use of materials for research purposes. All samples were obtained in accordance with the Declaration of Helsinki.

Competing interests

CS is a Founder, Chief Scientific Officer, consultant and shareholder of Amprion Inc., a biotechnology company that focuses on the commercial use of seed amplification assays for high-sensitivity detection of misfolded protein aggregates involved in various neurodegenerative diseases. The University of Texas Health Science Center has licensed patents to Amprion.