The microtubule-associated tau protein is an intrinsically disordered protein that play many different roles in cells [1–4]. In the brain, it is composed of four domains: The N-terminal domain, the proline rich region (PRR), the microtubule-binding region (MTBR) and the C-terminal domain. It can undergo alternative splicing of exon 2 and/or exon 3 that generates three N-isoforms (0N, 1N and 2N). Likewise, an alternative splicing of exon 10 located in repeat domain 2 (RD2) of the MTBR can occur, resulting in two R-isoforms (3R and 4R). In total, this results in six tau isoforms in the adult human brain[5]. Aside from different isoforms, tau is also known to undergo post-translational modifications (PTMs) such as phosphorylation, methylation, ubiquitination, glycosylation, acetylation and many more [6–9]. In pathological conditions, the tau protein can be found hyper- and abnormally phosphorylated. This has led to the hypothesis that these PTMs may be at the origin of a structural change in the tau protein that leads to its aggregation [10, 11].

Tauopathies is an umbrella term for a group of heterogeneous neurodegenerative brain disorders all characterized by the inclusion of tau aggregates within the cells. The tauopathy with the highest world prevalence is Alzheimer’s disease (AD), which is characterized by intracellular tau aggregates along with extracellular amyloid-ß plaques and therefore considered as a secondary tauopathy. Numerous primary tauopathies have also been described such as Progressive Supranuclear Palsy (PSP) and Pick’s disease (PiD) [12–14].

Although all tauopathies comprise aggregated tau, the tau lesions do not result in similar clinical deficits [15]. This clinical heterogeneity may be the consequence of the heterogeneity observed in numerous pathological aspects. First, the dominant isoforms of tau found within the aggregates are different leading to a post-mortem histopathological stratification of tauopathies. AD including both 3R and 4R tau is referred to as 3R/4R-tauopathy, while PSP is referred to as a 4R-tauopathy and PiD as a 3R-tauopathy [6, 16]. Secondly, the aggregates adopt different three-dimensional structures [17–19]. Thirdly, these lesions are described in different cell types of the brain [15]. Additionally, some tauopathies have a unique spatio-temporal staging based on the anatomical occurrence of tau lesions. For AD, these are described as the well-characterized Braak stages, which detail a tau pathology sequential evolution from the entorhinal cortex to the hippocampus, to finally invade the associative and primary neocortex [20–22]. A less robust hierarchical staging in PSP evolves from the pallido-luyso-nigral complex to the frontal and parietal lobes [23–26] and in PiD, it evolves through the limbic and frontotemporal neocortical regions [27]. The robust Braak stages in AD insinuate a cell-to-cell spreading of pathological tau, driving the prion-like tau propagation [28, 29]. For this, cytosolic translated tau is believed to be transported in the extracellular space through unconventional protein secretion (UPS) mechanisms, including extracellular vesicles (EVs) [30].

EVs are bilipid spheres known for their role in cellular communication through transport of proteins, lipids, RNA and more [31]. In the brain, EVs are directly released in the interstitial fluid surrounding neural cells and hence entitled brain-derived EVs (BD-EVs). Human BD-EVs of AD patients contain seed-competent tau [32–34]. Further, we demonstrated a different tau seeding capacity of AD, PSP and PiD patients-derived EVs. This reflects a heterogeneous contribution of BD-EVs to tau seeding among tauopathies, with AD BD-EVs showing the highest seeding capacity [34]. Here, we aim to identify and compare the tau proteoforms found within EVs of tauopathies to assess if this tau content lays at the base of the heterogeneity observed for EVs seeding among tauopathies. A tau immunoprecipitation followed by mass spectrometry analysis (tau IP-MS) performed on BD-EVs from non-demented controls (CTRL) or patients affected with different tauopathies, revealed detection of 22 tryptic tau peptides with two MTBR tryptic peptides enriched only in AD BD-EVs. Among them, the well-known pro-aggregative hexapeptide PHF6 (VQIVYK) was found [35–38]. Importantly, we demonstrated that PHF6-containing tau proteins are strongly involved in the seeding capacity of AD BD-EVs, where PHF6-ubiquitination is also part of it. From this, we conclude that the PHF6 is a driver for the higher EVs-mediated tau propagation in AD patients, revealing an interesting therapeutic target to prevent tau pathology spreading.

Human samples - The cohort in this study has been used in prior study published in Leroux and collaborators [34] and comprises of CTRL, AD, PSP and PiD prefrontal fresh-frozen brain extracts obtained from the Lille Neurobank (fulfilling French legal requirements concerning biological resources and declared to the competent authority under the number DC2008-642) with donor consent, data protection and Ethics Committee approval. Samples were managed by the CRB/CIC1403 Biobank, BB-0033-00030. A summary of the demographic data is listed in Table 1.

Brain-derived fluid isolation- To obtain the BDF, either papain or collagenase were used for the dissociation of the frozen human prefrontal brain extracts. Papain enzyme was used for brain dissociation to obtain the BD-EVs samples for tau-enriched mass spectrometry analysis. This was performed as previously described [39] and adapted in our previous study [34]. Briefly, frozen brain tissue (60 mg for proteomic analysis for each patient) was incubated on ice in Hibernate-A (50 mM NaF, 200 nM Na3VO4, 10 nM protease inhibitor (E64 from Sigma)) and then gently homogenized in a Potter before adding papain enzyme (2 mL of 20 units/mL; LS003119, Worthington). After a 20-minute (min) incubation at 37°C on wheel, 15 mL of cold Hibernate-A and protease inhibitor cocktail (Roche) were added and mixed by inversions to stop the enzymatic activity.

BD-EVs prepared for functional assays, involving immunodepletion (ID) and immunocapture (IC) with analysis of the seeding capacity on the fluorescence resonance energy transfer (FRET)-tau biosensor cells, were prepared using collagenase type 3 as this will only target the ECM and hence preserves EVs integrity [40]. The protocol using collagenase type 3 (LS0004182, Pan Biotech) for enzymatic tissue dissociation, was performed as previously described by Vella and collaborators [41]. Briefly, brain tissue (60 mg from a pool of three AD patients) was sliced on ice to generate smaller sections (~ 2 mm) before adding 75 units/mL of collagenase type 3 in Hibernate-E (800 µL per 100 mg of tissue, 10315538, Gibco). After an incubation of 20 min at 37°C with agitation, PhosSTOP (4906837001, Roche) and Complete Protease Inhibitor including EDTA (4693124001, Roche) were added to a final 1X concentration on ice.

For both protocols, successive centrifugations of 300 x g for 5 min, 2,000 x g for 10 min and 10,000 x g for 30 min at 4°C were applied to remove cells, membranes and debris, respectively. The final supernatant is entitled BDF and was consistently prepared fresh prior to each EVs isolation.

Brain-derived EVs (BD-EVs)- To isolate EVs from the BDF and to separate them from proteins contaminants, size-exclusion chromatography (SEC) was used as described previously by Leroux and collaborators [34]. Briefly, commercial SEC columns (ICO-70, IZON) packed with Sepharose resin CL-2B (CL2B300, Sigma-Aldrich) were equilibrated with degassed phosphate-buffered saline (PBS, 12559069, Gibco), 500 µL of BDF were applied on the SEC column followed by elution in degassed PBS. Once the void volume (3 mL) was eluted, the first 2 mL (F1-4) were recovered as EVs fraction in protein low binding tubes (0030108132, Eppendorf protein LoBind). We previously characterized this fraction enriched in BD-EVs in accordance with MISEV guidelines [34, 42].

One SEC was performed per individual patient for tau IP-Ms and from a pool of three AD patients (patient ID: F, K and M in Table 1) for ID and IC assays. The 2 mL EVs fraction were concentrated using ultrafiltration device 3 kDa Amicon (Amicon® Ultra-2 3 kDa, Millipore) at 4,000 x g (Multifuge X3R, Thermo Scientific) to a final volume of 50 µL/SEC/patient or 200 µL/SEC for IP-MS or ID/IC assays, respectively. Concentrated EVs were quantified by NTA.

Nanoparticle tracking analysis (NTA)- The concentration of particles was measured by NTA (NanoSight NS300, Malvern Panalytical) immediately after isolation. For this, BD-EVs samples were diluted in PBS and continuously infused into the NTA device by an automatic syringe pump at a flow rate of 20 µL/min. The focus was adjusted and the temperature was set to 25°C. Three videos of 60 seconds were acquired at camera level 15 and processed at detection level 4 using the NTA software [v 3.2.16]. Samples were freshly used for IC and ID experiments and only EVs designated for proteomic analysis were stored at -20°C before analysis.

Anti-tau antibody production and biotinylation- Antibodies used for the tau enrichment for IP-MS were home-made. Mice received injections every three-weeks of tau peptides with sequences 30–44 (for anti-Tau Exon 1 (TauE1C1)), 162–175 (for anti-TauPRR (TauP1)) or 427–441 (for anti-TauCter (Tau7F5)) coupled to KLH to amplify mouse immunisation. After four injections, lymphoid cells isolated from the spleen were immortalized by fusion with immortal myeloma cells SP2/0 in the presence of PEG/DMSO to generate hybridomas. Culture medium containing 15% of fetal bovine serum, 2 mM L-glutamine and 50 U/mL penicillin/streptomycin was used. For hybridoma selection, hypoxanthine aminopterin thymidine was added for 1 month following the protocol of Kohler and collaborators [43]. Clonal selection was done based on ELISA affinity to 1N4R recombinant tau protein. After two subcloning, clones were amplified. Monoclonal cells were cultured in CELLine flask (WCL1000-3, WHEATON) in Hybridoma-SFM (12045084, Gibco) complemented with 2 mM L-glutamine (25030081, Gibco) and 50 U/mL penicillin/streptomycin. Culture medium containing monoclonal antibodies was collected every 3 to 4 days and filtered using a Filtropur PES 0.45 µm membrane (83.1826, Sarstedt). The antibodies isotype was determined using the SBA ClonotypingTM System/HRP (5300-05, SouthernBiotech, 1/500 dilution in PBS) following manufacturers’ instructions and monoclonal antibodies were purified using affinity chromatography. For this, the automated ÄKTAprime plus (Cytiva) was coupled to a HiTrap HP column coated with protein A or G (17-0404-01 or 17-0402-01, Cytiva) and was used with the chromatogram profile obtained at absorbance 280 nm. First, column equilibration was done in filtered and degassed 20 mM sodium phosphate pH 7. A volume of 30 mL of supernatant from cell culture in CELLine flask with monoclonal antibodies was loaded and elution was done in 0.1 M glycine pH 2.7 where the pH of the antibody samples was directly neutralized in 1 M Tris buffer pH 9. Ten fractions of 500 µL were collected. Antibody containing fractions were selected after electrophoresis using a 1 hour (h) Coomassie brilliant blue staining (161–0406, Biorrad, 0.1% Blue G250 in 50% ethanol and 10% acetic acid) on denaturant NuPAGE gels (Thermofisher). Two subsequent dialysis of selected antibody fractions were done for storage in PBS. Antibody concentrations were measured using the NanoDrop™ One (ThermoFisher) at 280 nm. Antibody epitopes were verified by nuclear magnetic resonance (NMR) by comparing spectra of ¹⁵N 2N4R tau in presence and absence of the antibody. These results along with the validation of detection of human tau for the three home-made tau antibodies is shown in supplementary Fig. 1.

The produced TauE1C1, TauP1 and Tau7F5 antibodies were biotinylated using the EZ link Sulfo-NHS-SS-biotinylation kit (21445, Thermo Fisher) following the protocol recommendations of the manufacturer. Briefly, 2 mg antibody in 1 mL filtered PBS were incubated with 10 mM biotin on a wheel at low speed at 4°C for 2h in protein low binding Eppendorf. After this, the biotinylated antibody was separated from unbound biotin using 5 mL Zeba Spin Desalting columns (7K MWCO, 89981, Thermo Fisher). The eluate containing the antibody-biotin complex was collected and concentrations were measured using the NanoDrop™ One (Thermo Fisher). Biotinylation was validated using ELISA against recombinant 1N3R tau protein (Supplementary Fig. 2A-B). Biotinylated antibodies were stored at 4°C.

¹⁵ N-labeled recombinant tau- ¹⁵N-labeled (entitled, heavy or isotopically labeled) recombinant tau0N4R, tau1N3R and tau2N4R were generated as described previously by Danis and collaborators [44]. Briefly, the pET15b-Tau recombinant T7 plasmids were transformed into competent E. coli BL21 (DE3) bacterial cells of which a small-scale culture was grown in Luria-Bertani medium at 37°C. For production of recombinant ¹⁵N-labeled tau0N4R, tau1N3R and tau2N4R, the small-scale culture was cultured in 1 L of a modified M9 medium containing MEM vitamin mix 1 × (Sigma-Aldrich), 4 g of glucose, 1 g of ¹⁵N-NH₄Cl (Sigma-Aldrich), 0.5 g of ¹⁵N-enriched Isogro Growth Powder (Sigma-Aldrich), 0.1 mM CaCl₂, and 2 mM MgSO₄. Once the culture reached an optical density of 0.8 at 600 nm, 0.5 mM isopropyl-β-D-thiogalactopyranoside was added to induce expression of ¹⁵N-recombinant tau. Next, the cells were lysed in 50 mM NaPi pH 6.5 with 2.5 mM EDTA, complete protease inhibitor cocktail (Sigma-Aldrich). The tau proteins were first purified by heating the bacterial extract for 15 min at 75°C using a water bath. After centrifugation, the resulting supernatant was then passed on a cation exchange chromatography column (CEX, Hitrap SP Sepharose FF, 5 ml, Cytiva) equilibrated in 50 mM NaPi pH 6.5 and eluted with a NaCl gradient. A buffer exchange with 50 mM ammonium bicarbonate (Hiload 16/60 desalting column, Cytiva) of the recombinant tau proteins was done before lyophilization. ¹⁵N-labeled recombinant tau was suspended to 1 mg/ml in 50 mM ammonium bicarbonate and 1 mM bovine serum albumin. The solutions were diluted at 100 µg/mL and then aliquoted into LoBind tubes and stored at -80°C until use to avoid thawing cycles.

BD-EVs tau enrichment and proteomic sample preparation- BD-EVs samples (2x10¹⁰ particles) from individual patients were incubated with MB2 solution (40 mM Tris-HCl, 137 mM NaCl, 2 mM EDTA, 10 mM Guanidine and 2% IGEPAL), 20 ng of ¹⁵N-labeled recombinant tau (8 µg 2N4R, 6 µg 1N3R and 6 µg 0N4R) and a combination of three home-made biotinylated monoclonal anti-tau antibodies (2 µg TauE1C1, 1 µg TauP1 and 2 µg Tau7F5) overnight at 4°C while shaking. The added value of tau enrichment using a combination of three tau antibodies, targeting distinct regions of the protein, is demonstrated in Supplementary Fig. 2C-D. Tau immunocapture was performed by the AssayMAP Bravo platform (Agilent Technologies) using Steptavidin cartridges and the "Affinity purification" program. The sample was eluted with 25 µL of elution buffer (12 mM NaCl, 100 mM HCl, pH 2) and then neutralized with 20 µL of 100 mM Tris-HCl at pH 8.5. The eluate was digested in a mix of 1M urea for denaturation, 10% acetonitrile (ACN) and 0.5 µg Trypsin-LysC mix (Promega) for 6 h at 37°C at 450 rpm. The digestion was stopped by addition of 2% trifluoroacetic acid. A desalting of tryptic peptides was performed by the AssayMAP Bravo platform (Agilent Technologies) using C18 cartridges and the "Peptide Clean-Up" program. Samples were eluted with 30 µL of elution buffer (40% ACN, 0.1% formic acid (FA)). Sample were dried at 50°C for 1 h with a SpeedVac instrument (LabConco) and were suspended in 20 µL of the mobile phase A (0.1% FA in water; Biosolve).

LC-MS/MS analysis- Tryptic peptide separation was performed using an EvoSep One liquid chromatography system (EvoSep) with a PepSep C18 column (15 cm x 150 µm, 1.5 µm, Bruker Daltonics). A 34 min elution gradient was used corresponding to 30 Sample Per Day (SPD) using mobile phase A (0.1% FA in water; Biosolve) and mobile phase B (0.1% FA in acetontrile; Biosolve). Peptides were analyzed using a trapped ion mobility spectrometry quadrupole time-of-flight mass spectrometer (timsTOF HT, Bruker Daltonics) equipped with a nano-electrospray ion source (Captive spray, Bruker Daltonics) in positive-ion mode. Data were acquired in a Data-Dependent Acquisition (DDA) mode with a TIMS accumulation time of 100 ms and a number of PASEF ramps of 10. The PASEF scan mode was performed in the mass range 10-1700 m/z and ion mobility range 0.75–1.25 1/Ko.

Proteomic data analysis- Raw files were processed with Skyline-daily software 24.1.1.398 (MacCoss Lab) with following parameters: centroided precursor mass analyzer: MS1 mass accuracy of 10 ppm, centroided product mass analyzer: MS2 mass accuracy of 10 ppm, include all matching scan. The extracted ion chromatograms of selected fragments were manually reviewed and peak picking adjustments were made when needed. The most intense of precursor ion peak area was calculated in Skyline and exported to Excel (Microsoft) for subsequent statistical analysis. To define the tryptic peptide range, all tryptic peptides were aligned on the full-length 2N4R isoform (441 AA).

The peak areas were normalized based on the peak areas of heavy ¹⁵N-labelled tau standard peptides when possible and corrected for the molar distribution of exon 2, exon 3 and exon 10 heavy ¹⁵N-labelled tau added. The sum of normalized peak areas is taken for tryptic peptides with a missed cleavage, namely [210–224] and [354–370]. Some oxidations on methionine were detected and considered as artefacts as they may have been induced during sample preparation.

As ¹⁵N-labelled recombinant tau do not contain PTMs, FragPipe software v22 was used to determine potential tryptic peptides with PTMs of interest such as phosphorylation and ubiquitination. Raw files were processed with MSFragger v4.1 to search for PTMs with serine, threonine and tyrosine phosphorylation and lysine ubiquitination as variable modifications. The database was downloaded directly by the platform from Uniprot Swiss-Prot with protein isoforms, reverse sequences and contaminants. Search parameters were set as follows: trypsin as protease, 2 miss-cleavages, 2 as maximum number of variable modifications, 7–30 amino-acids peptide length range, 20 ppm precursor and fragment mass tolerance. Only peptides with a peptide-spectrum match score above 0.99 were retained.

Variable domain of the Heavy chain of the Heavy chain only antibodies (VHH) production and biotinylation- To define seed competent tau species enriched in AD EVs, VHH targeting the human tau protein were employed. More specifically, VHH Z70 [305–312] and VHH A5-2 [330–370] were used [45, 46]. A VHH anti-green fluorescent protein (GFP) was used as control. These VHHs used for the IC of tau inside EVs were produced and purified as published in Danis and collaborators and contain a C-terminal cysteine [45]. Briefly, periplasmic extracts of Escherichia coli BL21 (DE3) bacterial cells were retrieved after transformation with various pET22b VHH-cystein and pET22b VHH without C-terminal tag constructs [46]. The VHHs were purified by immobilized-metal affinity chromatography (IMAC, HisTrap HP, 1 mL, Cytiva) followed by SEC (Hiload 16/60, Superdex 75, prep grade, Cytiva) in phosphate buffer (50 mM sodium phosphate buffer [NaPi], pH 6.7, 30 mM NaCl, 2.5 mM EDTA, 1 mM DTT). VHHs were dialyzed against 50 mM Tris pH 8, 50 mM NaCl and cleaved with His-tagged Tobacco Etch virus (TEV) protease. The TEV protease and the cleaved 6-His tag were removed by a second IMAC step, and the VHHs were recovered in the flow-through, concentrated and flash-frozen for further use. A buffer exchange was done to obtain VHHs in PBS (ET330, Euromedex). All recombinant VHHs with a cysteine at the C-terminus were biotinylated with 10 molar excess of maleimide biotin conjugates (A39261, Thermo Fisher Scientific) overnight at 4°C. The reactions were quenched with 5 mM DTT. Residual biotin conjugates were removed by two consecutive buffer exchanges using a Zeba Spin Desalting column (7K MWCO, 89882, Thermo Fisher Scientific). The eluate containing the VHH-biotin complex was collected and concentrations were measured using the NanoDrop™ One (Thermo Fisher). Biotinylated VHHs were stored at -80°C after flash freeze in liquid hydrogen.

Anti-ubiquitin antibody biotinylation- To perform the single and double IC of ubiquitinated tau, the commercially available mono- and poly-ubiquitinated conjugates recombinant monoclonal antibody (UBCJ2) (ENZ-ABS840, immunoglobulin-1 (IgG1) isotype, ENZO) and a mouse IgG1 isotype control antibody (02-6100, Thermo Fisher) were biotinylated using the EZ link Sulfo-NHS-SS-biotinylation kit (Thermo Fisher, 21445) following the protocol recommendations of the manufacturer. Briefly, 0.4 mg antibody in 0.1 mL filtered PBS was incubated with 10 mM biotin on a wheel at low speed at 4°C for 2 h in protein low binding eppendorf. After this, the biotinylated antibody was separated from unbound biotin using 0.5 mL Zeba Spin Desalting columns (7K MWCO, 89882, Thermo Fisher). The eluate containing the antibody-biotin complex was collected and concentrations were measured using the NanoDrop™ One (Thermo Fisher). Biotinylated antibodies were stored at 4°C.

EVs tau ID and IC- To assess the seeding capacity of particular tau peptides inside AD-derived EVs, an IC and ID were performed. For this, BD-EVs derived from a pool of three AD patients were prepared freshly and sonicated for 30 min in a waterbath at 200 W with 30 sec intervals (Bioruptor® Sonication System, Diagenode) to which ice was added on regular bases.

In parallel, the IC using the Cytiva streptavidin Mag Sepharose magnetic beads (28985799, Cytiva) was performed as follow. 50 µL streptavidin coated magnetic beads in protein low binding eppendorfs were washed in 500 µL of binding buffer (Tris-buffered saline pH 7.5). Next, 60 µg biotinylated VHH (VHH Z70, VHH A5-2 or VHH anti-GFP) or 5 µg antibodies (anti-Ubiquitin or anti-IgG1 isotype) in 300 µL binding buffer were added to the beads and incubated for 30 min at room temperature (RT) on a rotor (Grant bio PTR-35) using end over end rotation with intermittent vibrations to ensure successful homogenization. After incubation, unbound VHH or antibody were removed, and two washes were done using 500 µL binding buffer. Then, 300 µL binding buffer containing 1x10¹⁰ AD-EVs (for VHHs-based IC/ID) or 4x10¹⁰ AD-EVs (for antibody-based IC/ID) were added. This mixture was incubated for 1 h at RT on a rotor (Grant bio PTR-35) using end over end rotation with intermittent vibrations to ensure successful homogenization. After this, samples were placed on the magnetic rack. The medium was removed and kept as ID fraction containing all proteins except the tau with the sequence of interest. These 300 µL of ID fraction were concentrated using 3K amicon to a final volume of 100 µL.

For the elution of VHH-bound tau or antibody-bound tau, 60 µL of 0.1 M glycine-HCl at pH 2.5 were added to the beads for 2 minutes while tubes were gently shaken. The pH was neutralized by addition of 4 µL of 1 M tris base pH 11 buffer. After mixing and placement on the magnetic rack, the solution containing tau previously captured by the VHH or antibody, called the IC fraction, was collected and the volume was adjusted to 100 µL using PBS.

For the double IC, the IC fraction obtained using anti-ubiquitin was diluted to 300 µL and added on the magnetic beads coupled to the VHH for a second incubation of 1 h. This was followed by collection of the ID fraction and elution of the IC fraction as described above.

The ID and IC using VHH Z70 and anti-ubiquitin antibody of tau from AD brain homogenate were validated by western blot (Supplementary Fig. 3–4). The enrichment of BD-EVs tau peptides using VHH Z70 or VHH A5-2 have also been validated by mass spectrometry analysis as shown in supplementary Fig. 5.

Cell culture- The stable Tau RD P301S FRET Biosensor cells (ATCC CRL-3275) and HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, 13345364) with pyruvate and without HEPES complemented with glutaMax 1X (Gibco-35050061), 10% fetal bovine serum (Gibco, A5256701) and 1% penicillin-streptomycin. The cells were maintained in a humidified incubator with 5% CO₂. Cell splitting was done twice a week.

FRET-tau biosensor cell assay- HEK FRET and HEK 293T cells were plated into a 12-wells plate (150,000 cells per well) 24 h before treatment. Sonicated Tau [244–368], also called K18 fibrils (2 µM, as prepared in Danis and collaborators [45]) were used as a positive control and PBS as a negative control. The ID and IC fractions were lipofected onto the cells. For this, the 100 µL ID or IC fraction was completed with 100 µL of Lipofectamine2000 (Life Technologies, 11668019) diluted at 1:10 in Opti-MEM (Fisher Scientific, 31985062). These transfection mixtures were incubated for 20 min at RT and gently added in the culture medium of the cells. After 72 h, the cells were collected in warm PBS and the Zombi NIR labelling, and fixation protocol were performed as previously described in Leroux and collaborators [34]. The cells were analyzed on the flow cytometer Aria SORP (BD Biosciences; acquisition software FACSDiva v7.0, BD Biosciences) or the Spectral SONY cytometer (SP6800) with the following excitation/emission wavelengths: excitation 405 nm, CFP emission 466 ± 40 nm and FRET YFP 529 ± 30 nm; excitation 488 nm, YFP emission 529 ± 30 nm were used. The FRET data were quantified using the Kaluza analysis software v2. VHH IP experiments were repeated six times, single anti-ubiquitin IP experiments were repeated three times and double IP experiments were repeated six times, all with at least 10,000 cells per replicate analyzed.

Statistical analyzes- Statistics and graphs were generated using GraphPad Prism 9 software (version 9.1.0). Data were represented as mean ± standard error of mean (SEM). Shapiro-Wilk normality test was used to define normality of each group. Comparison of two independent groups with a normal distribution were done using a t-test and for groups with non-parametric distributions, a Mann-Whitney U test was used. Comparisons of three or more independent groups, with a normal distribution were done using ordinary one-way Analysis Of Variance (ANOVA), while Kruskal-Wallis test was used for non-parametric samples. Statistical testing was done at the two-tailed p-value of 0.05.

Numerous tau peptides were detected within BD-EVs from CTRL, AD, PSP and PiD- The tau seeding capacity of BD-EVs from different tauopathies indicates an important heterogeneity [34]. This may be explained by the tau proteoforms and hence, prompt our decision to assess the tau profiles found within the EVs of CTRL, AD, PSP and PiD patients. For this, enzymatic brain dissociation of 4 CTRL, 10 AD, 9 PSP and 4 PiD brain samples was followed by SEC for EVs isolation. The EVs quality and characteristics have already been published by Leroux and collaborators [34]. This was complemented with a global proteomics using a data independent acquisition mode (DIA) which also showed a significant enrichment in EVs-associated proteins compared to proteins known as EVs contaminants (Supplementary Fig. 6).

To unravel the tau proteoforms present within BD-EVs by mass spectrometry, a tau-IC was set-up. The combination of three home-made tau antibodies (TauE1C1, TauP1 and Tau7F5) with epitopes across the entire tau sequence were selected to enable the best tau sequence recovery (Fig. 1A, Supplementary Fig. 2C-D). For normalization and hence, comparison among patients and tauopathies, heavy ¹⁵N-labelled recombinant tau was added along with the antibodies to the BD-EVs. After incubation, a streptavidin-based purification was done followed by trypsin digestion of tau before mass spectrometry analysis (Fig. 1B). In total, 22 peptides of tau, 4 phosphorylated tau peptides and 1 ubiquitinated tau peptide were detected in BD-EVs (Fig. 1C-D). These detected peptides are also represented in a size-scaled alignment on full-length 2N4R tau (Fig. 1E). This tau IP-MS with a mix of three antibodies resulted in a 57.6% tau sequence recovery with the highest (89.8%) recovery in the N-terminal domain (Supplementary Fig. 7). The MTBR and C-terminal domain were found to have a 49.6% and 42.5% sequence recovery, respectively.

Tau profiles observed between BD-EVs from CTRL, AD, PSP and PiD reveal AD-enriched peptides- To start, a comparison of tau peptides among BD-EVs of tauopathies was done (Fig. 2). This shows enrichment in PSP compared to PiD for some peptides in the N-terminal region, PRR and C-terminal region. Interestingly, the 3R [306–317], [354–370] and [376–383] peptides showed a significant enrichment in AD-derived EVs compared to PSP. This last one is also found significantly enriched in AD compared to PiD. Also, peptide [386–395] is significantly enriched in PSP-enriched EVs compared to PiD and AD; peptide [396–406] is more represented in PSP-derived EVs in comparison to PiD. These data clearly indicate heterogeneity in the proteoforms of tau among tauopathies.

Further, peptides were aligned on the 2N4R tau sequence (Fig. 3A). Interestingly, two MTBR peptides namely: 3R [306–317] and [354–370] were significantly enriched in AD BD-EVs (Fig. 3B).

PHF6-containing tau in AD BD-EVs is an actor of the EVs-mediated seeding- The peptides of the MTBR have previously been shown enriched in AD sarkosyl-insoluble fractions [47, 48] and recently, Fowler and collaborators found MTBR peptides enriched in AD-EVs as we demonstrate here [33]. Interestingly, one of the AD-enriched peptides, namely 3R [306–317] comprises the PHF6 (VQIVYK) hexapeptide, which is a sequence implicated in tau aggregate formation [35–37]. Hence, we investigated whether these AD-enriched peptides transported within EVs explain the high tau seeding capacity observed only in AD patients. To answer this question, immunprecipitations of these tau peptides in AD BD-EVs were performed using single-domain antibodies: nanobodies®, also known as VHHs. We developed and characterized VHHs directed against different tau epitopes [45, 49–51]. We chose two anti-tau VHHs: one (Z70) targeting the PHF6 sequence of the R2 domain and the other the [354–370] sequence of the R4 domain of tau (Fig. 4A). A VHH targeting GFP was used as a negative control. After sonication, AD BD-EVs were incubated with each VHH coupled to magnetic beads (Fig. 4B). Both IC and ID fractions were analyzed. IC fractions were transfected onto HEK FRET-tau biosensor cells. The FRET-positive percentage of IC fractions was normalized against anti-GFP ID (theoretically containing all seeding species) set at 100% (Fig. 4C-D). Our results show a significant strong increase in tau seeding by PHF6-containing peptides (VHH Z70 IC) and a mild but significant increase in tau seeding by [354–370] epitope-containing peptides (VHH A5-2 IC) compared to anti-GFP VHH. The Z70 IC therefore confirms the best enrichment of pro-nucleating species and the A5-2 IC indicates that some pro-nucleating tau species also include the R4 domain.

Similarly, ID fractions containing tau peptides unbound to VHH A5-2 or Z70 were collected and transfected onto HEK FRET-tau biosensor cells. As for IC, the ID fraction with anti-GFP VHH containing all tau seeds was fixed at 100% and used for normalization of the other ID fractions (Fig. 4D). Our results show that the ID fraction of tau peptides lacking the PHF6 sequence (VHH Z70 ID) in AD BD-EVs has a significantly decreased tau seeding capacity of 61% (Fig. 4D). In contrast, the remaining tau peptides in the VHH A5-2 immunodepleted fraction retained tau seeding capacity. Altogether, these results confirm that among the VHH-targeted tau peptides enriched in AD BD-EVs, tau peptides containing the PHF6 sequence are the most involved in EV-mediated tau seeding.

Ubiquitinated tau peptides detected inside BD-EVs of AD- Given the significant recovery of tau peptides, we further analyzed these peptides for disease-relevant PTMs, which play a pivotal role in tau pathology. Phosphorylation is the best-studied tau PTM [6]. Here, a total of four phospho-peptides (pT181, pS202, pT231 and pT403/pT404) were detected inside BD-EVs among tauopathies, with the majority located in the PRR (Fig. 1E). Of these, pT181 was shown significantly enriched in BD-EVs of CTRL compared to PiD, and pT403/pS404 was found enriched in BD-EVs of CTRL compared to AD (Supplementary Fig. 8). Further, no tau acetylation was detected. Nevertheless, tau is a lysine-rich protein and is therefore susceptible to ubiquitination, which regulate tau clearance [8, 52]. Our tau-enriched IP-MS did reveal ubiquitination on the 3R [306–317] peptide (Ub-PHF6) only in AD-EVs (Fig. 5A). Interestingly, this is located on the PHF6 peptide that we previously shown to have a high implication in EVs-mediated seeding (Fig. 4C-D). To investigate whether pro-nucleating species of tau may be ubiquitinated, we performed anti-ubiquitin immunoprecipitation on AD BD-EVs and analyzed IC and ID fractions (Fig. 5B). An anti-IgG1 was used as an isotype control. The ID and IC fractions were transfected in the FRET-tau biosensor cell assay. The ID fraction with the IgG1 isotype theoretically containing all AD BD-EVs tau peptides was set to 100% for normalisation of all ID and IC conditions (Fig. 5C). As expected, the IgG1 IC fraction did not contain any pro-seeding tau species. In contrast, the IC Ub fraction showed tau seeding capacity, but this remained relatively low at around 15%. For the ID fractions, fraction ID IgG1 includes pro-nucleating tau species. Fraction ID Ub contains less ubiquitinated tau, revealing a still high tau seeding capacity. To confirm the mass spectrometry results (Fig. 5A), we immunoprecipitated tau peptides containing the PHF6 sequence from the IC Ub fraction using VHH Z70 and analyzed the resulting IC and ID fractions for their ability to induce tau seeding. Anti-GFP VHH was used as a control (Fig. 5B). The IC and ID fractions from the double immunoprecipitation were transfected onto the FRET-tau biosensor cell assay. The ubiquitinated positive ID fraction obtained after anti-GFP (IC Ub + anti-GFP ID) was set to 100% for normalisation (Fig. 5D). The results indicate that Ub-PHF6 tau species (IC Ub + IC Z70) in AD-derived EVs have a higher seeding capacity than anti-GFP VHH controls (Fig. 5D, left). Selective removal of ubiquitinated tau protein containing PHF6 (IC Ub + ID Z70) did not result in a significant reduction in seeding capacity (Fig. 5D, right) suggesting that ubiquitinated forms of PHF6 tau peptides may induce pro-seeding activity but it remains marginal.

Discussion- Tauopathies, although all characterized by the presence of tau lesions within the brain, are heterogeneous based on their clinical symptoms, tau filament cryo-electron microscopy (EM) conformation and isoform, affected cell type, spatio-temporal staging and many more [15]. Recently, it was demonstrated that BD-EVs -mediators of the intercellular transfer- have different seeding capacities among tauopathies, with AD EVs having the highest seeding capacity [33, 34, 53]. This could be related to the nature of tau species transported within BD-EVs. At this day, only one study has identified the tau proteoforms of AD BD-EVs [33]. In this article, we went further and identified the tau profiles present within the BD-EVs from AD, PSP, PiD and CTRL. For this, we performed an IC of tau coupled to mass spectrometry analysis (IP-MS), revealing detection of 22 tryptic tau peptides in all conditions and without phosphorylation or ubiquitination, with a total sequence recovery of 57.6%.

Studying tauopathies, we assessed if the dominant isoform found within tau filaments of tauopathies is reflected in the isoform distribution inside EVs (Supplementary Fig. 9). A consistent correspondence was found only for the 4R-tauopathy PSP, having significantly more 4R isoform present within BD-EVs. Previous proteomic studies indicated that isoform distribution of sarkosyl-insoluble fractions correlate with the dominant isoform in filaments of tauopathies [48, 54]. Hence, our findings suggest that BD-EVs from tauopathies do not necessarily mirror the isoform composition of the aggregates. Ruan and collaborators studied the BD-EVs of AD patients and found enrichment of oligomeric tau compared to control BD-EVs [53]. Nevertheless, some paired helical filaments (especially in very large-sized EVs) have been detected in BD-EVs of AD [33, 53], and therefore it is of interest to map the relative proportion of oligomeric tau and filaments within BD-EVs and to assess their contribution to tau seeding. To fully map these different tau structural assemblies (truncated tau, oligomeric tau, filaments…) within BD-EVs from tauopathies, cryo-EM, high- performance liquid chromatography (HPLC)-coupled SEC or top-down proteomics is required in the future.

The tau IP-MS showed peptides of the MTBR RD3 and RD4 significantly enriched in BD-EVs of AD compared to CTRL, as equally observed by Fowler and collaborators [33]. MTBR has also been described in the cerebro-spinal fluid and in sarkosyl-soluble and sarkosyl-insoluble fraction of AD patients [48, 55]. Enriched MTBR peptides in AD-EVs may explain why these vesicles are more seed competent than EVs from other tauopathies. Therefore, IC and ID were performed to study the seeding capacity of these two AD enriched tau peptides using two different VHHs: VHH Z70 (anti-RD3) and VHH A5-2 (anti-RD4). The ID VHH A5-2 and VHH Z70 showed 27.8% and 50.8% reduction of the targeted peptides, respectively. IC VHH A5-2 had a capture efficiency of 1.4%, and VHH Z70 of 29.6% measured by mass spectrometry (Supplementary Fig. 5). The equilibrium dissociation constant (K_D) of VHH A5-2 (K_D = 576 nM) was approximately two times higher than VHH Z70 (K_D = 246 nM) meaning more tau dissociation occurs for VHH A5-2. Also, the Koff values are different (koff = 56.3 x 10^− 3 s-1 for A5-2 and 3.53 x 10^− 3 s-1 for Z70 (Supplementary Fig. 10)) with Z70 Koff 16 times lower than A5-2 [46]. Together these VHH parameters could explain these results. By transfecting ID and IC VHH Z70 fractions on the FRET-tau biosensor cell assay, we demonstrated that the enriched peptides in AD-EVs containing the PHF6 hexapeptide have a high seeding capacity (Fig. 4C-D). The low capture efficiency combined with the high seeding capacity of tau containing the IGSLDNITHVPGGGNKK [354–370] sequence suggests that this tau specie is also implicated in EV-mediated seeding although we cannot exclude that it also contains the PHF6 sequence (Fig. 4C-D). Multiple research teams have demonstrated that the PHF6 (VQIVYK) hexapeptide is an active core for the templated misfolding[35, 36, 56–58]. Recently, Lövestam and collaborators visualized the implication the PHF6 as forming the first intermediary amyloid filaments, acting as the primary nucleation of recombinant tau by cryo-EM [37]. Our data clearly demonstrated that, in addition to its pro-aggregative role inside the neuron, the PHF6 motif is also a seed competent peptide implicated in human EVs-mediated pathology progression. In the future, it would be of interest to map at which Braak stage, the enriched abundance of PHF6-containing tau in the BD-EVs occurs to understand the kinetics of propagation via EVs.

The tau protein can undergo numerous PTMs, which has driven research to map disease-specific PTMs. Studying the sites of tau phosphorylation, we detected only four phospho-peptides and found a decrease in pT403/pS404 tau in AD-EVs compared to CTRL (Supplementary Fig. 8). There is a physiological phosphorylation of tau proteins and post-mortem tau dephosphorylation⁶ may not be as efficient than in neurons. However, other research groups have shown in AD-EVs the presence of pS404 tau while using the conformational dependent PHF-1 antibody for western blot [33, 53]. The observed differences are likely technique dependent as mass spectrometry allows detection of tau phospho-peptides in a conformation independent manner allowing of numerous phosphorylations in CTRL BD-EVs. The four detected phospho-peptides are known to be phosphorylated in both physiological conditions and in AD 6,59. Interestingly, they are also detected in BD-EVs of PSP and PiD. Our results suggest a basal transport of phospho-tau species within EVs, although the majority of tau transported within EVs is non-phosphorylated.

Based on our tau IP-MS, we detected ubiquitination on the PHF6-containing tryptic peptide (3R-specific) unique to AD-EVs. This Ub-PHF6 has been previously described to be enriched in AD patients brain lysate (soluble and insoluble fractions) [54, 60], but this is the first demonstration that it is also present within AD-EVs. As we showed a crucial role of the PHF6 tau species inside EVs for tau seeding, we assessed if Ub-PHF6 tau species from EVs present a seeding capacity. Our results from the single IC using an anti-ubiquitin antibody and the double IC using an anti-ubiquitin antibody followed by VHH Z70, indicate that ubiquitinated tau species from EVs are capable to induce tau nucleation. Considering the capture efficacies using both antibody anti-ubiquitin and VHH Z70 (Supplementary Fig. 3–5), only a small fraction of Ub-PHF6 tau Ub-PHF6 tau species were isolated. However, these tau species from AD BD-EVs proved to be seed-competent. Previously, research has shown that ubiquitination can enhance tau aggregation assembly, and it was hypothesized to be related to the neutralization of the MTBR positive charge [61–64]. The occurrence of ubiquitinated tau inside EVs, drives the hypothesis that affected cells have saturated ubiquitin-proteasome degradation systems and hence try to discard this pathological tau via encapsulation in EVs for degradation in other brain cells (incl. microglia or healthy neurons) [65, 66]. Another explanation might be that some PTMs affect the addressing of protein to degradation systems.

Conclusions- Our observations highlight that the tau profile within BD-EVs is different among tauopathies, with enrichment of PHF6-containing peptide in AD BD-EVs. We demonstrate a crucial role of the 3R-specific PHF6-containing tau, in EVs-mediated tau seeding of AD. This insight advances our understanding of AD mechanisms by showing a sorting process that favors PHF6 seeding-competent tau into EVs. Hence, the cellular machinery possibly contributes to loading toxic tau conformers into EVs driving to disease progression.