Why do some people with AD slide rapidly into severe dementia, while others decline gradually over more than a decade? Part of the answer could come down to which biochemical forms of tau inhabit a person’s brain, suggests a study published June 22 in Nature Medicine. Among postmortem brain samples from people with advanced AD, Bradley Hyman at Massachusetts General Hospital in Charlestown and colleagues found a striking variability in tau’s ability to seed aggregation. The scientists tied aggregation-prone forms of tau in the postmortem brain to a more rapid course of disease during life. They pegged large, soluble tau oligomers—phosphorylated on specific residues—as the most hazardous species. Antibodies trained against these types of tau stopped its aggregation. The findings cast tau’s behavior as a major prognostic determinant for AD, and support the concept of targeting these troublesome forms of the protein with therapeutics.

  • Tau seeding activity varied by an order of magnitude among 32 people with AD.
  • Seeding activity, hyperphosphorylation, and oligomerization of tau correlated with clinical aggressiveness.
  • Phosphorylation of specific tau residues associated with rate of decline.

“This is a well-designed study underlining once more the importance of the soluble tau oligomeric assemblies over long tau filaments in Alzheimer’s disease, as well as heterogeneity in tau oligomers,” commented Rakez Kayed of the University of Texas Medical Branch in Galveston.

“The study clearly highlights the complexity and heterogeneity of tau proteins in people with AD,” noted Hilal Lashuel of École Polytechnique Fédérale de Lausanne in Switzerland. The complexity calls for cautious interpretation of the findings, Lashuel added, noting that enigmatic tau oligomers are in dynamic equilibrium that can be influenced by disease stage and all manner of other factors. “It would be very difficult to identify a specific oligomer species that consistently correlates with tau propagation or disease progression,” he said.

Throughout the course of AD, neurofibrillary tangles of tau overtake the brain in a stereotypical sequence (Braak and Braak, 1991). Fueled at least in part by a templated misfolding mechanism, the propagation of tau tangles throughout the brain is tied closely to clinical progression of the disease (Jan 2020 news; May 2019 newsJun 2019 news). 

But even among people with the typical, amnestic form of AD, how aggressively their clinical disease gets worse varies strikingly from one person to the next (Komarova et al., 2011). This not only creates uncertainty for patients and their families, but also poses a risk for clinical trials, whose success depends on being able to measure a treatment effect within a set time, typically six months for Phase 2. Too many slow progressors in a trial cohort can sink a study even if the drug did what it was intended to do. Might molecular variations in tau species—particularly those that influence its propagation—explain this clinical heterogeneity?

Diversity of Decline. CDR-SOB scores worsened (increased) at vastly different rates in people who ultimately died with advanced AD. Some never reached the maximum score before they died. [Courtesy of Dujardin et al., Nature Medicine, 2020.]

First author Simon Dujardin and colleagues addressed this question by probing myriad aspects of tau taken from the postmortem brains of 32 people who had died in the advanced stages of AD. At the time of death, each had extensive tau tangles in the brain, at Braak stage V/VI. However, their clinical trajectories had been remarkably variable. Their age at onset ranged from 45 to 81, and the time between symptom onset and when they died ranged from five to 19 years. Their rates of cognitive decline, as gauged by serial tests on the clinical dementia rating scale sum of boxes (CDR-SOB), also varied widely.

The researchers started by measuring the capacity of tau to seed aggregation in biosensor cell lines. Equipped with a tau fragment tagged with fluorescent donor and acceptor molecules, these cells light up when potent seeds spark aggregation (Oct 2014 news). Although the researchers normalized the amount of tau used from each brain, they found wide variation in seeding activity. It ranged by an order of magnitude across samples. Notably, the three samples with the highest seeding activity came from people who carried two copies of ApoE4, suggesting the risk factor influences tau propagation.

Seed Span. When added to biosensor cell lines (left), soluble tau proteins extracted from the brains of people with AD were strikingly heterogeneous in a seeding assay. [Courtesy of Dujardin et al., Nature Medicine, 2020.]

The physiological relevance of this cell-based biosensor assay has been challenged recently (May 2020 news). So as not to rely entirely on this assay as a proxy for tau’s propagation potential, the researchers also conducted a series of alternative seeding experiments with a subset of the samples deemed low, intermediate, or high seeders based on the biosensor assay. Whether treating primary neuron cultures with these extracts or injecting the extracts directly into the brains of mice expressing P301S tau, the researchers observed similar relative trends in seeding activity among the samples. This suggested that the cell-based biosensor assay provided a meaningful gauge of the relative potency of tau seeds in each sample.

Planting the Seed. Tau extracted from human brain samples with low, moderate, or high seeding activity on biosensor assays triggered similar trends of tau aggregation when injected into the P301S mouse brain. [Courtesy of Dujardin et al., Nature Medicine, 2020.]

To figure out what about tau determines its seeding potency, the researchers subjected tau in the brain extracts to a barrage of biochemical assays and cross-referenced the results with the seeding activity gleaned from biosensor assays. They report that seeding activity correlated not with a person’s amount of total tau, but with levels of oligomeric, hyperphosphorylated tau in each sample. Compared with intermediate or low "seeders," high seeders had an abundance of soluble, high-molecular-weight tau oligomers.

Using mass spectrometry, the researchers mapped the phosphorylation landscape of tau across samples, noting that tau doubly phosphorylated on Thr231 and Ser235, or singly phosphorylated on Ser262, correlated with seeding activity. Curiously, neither ptau-181 nor ptau-217—the species that rise in the cerebrospinal fluid in the preclinical stages of AD—were significantly tied to seeding activity (Mar 2020 news; Apr 2020 conference news).

Does tau seeding activity, or any of its biochemical correlates, relate to a patient’s clinical progression? Indeed, the researchers found that the higher the tau seeding activity, the steeper the person’s rate of decline on the CDR-SOB, and the younger his or her age at symptom onset. The abundance of oligomeric, hyperphosphorylated species of tau also correlated with disease progression, as did levels of the same phospho-tau species that associated with seeding activity.

Collectively, tau seeding activity accounted for about 25 percent of the clinical heterogeneity among people with typical AD, the researchers reported. This suggests that tau antibodies that block tau seeding might also stem clinical progression of the disease. To identify antibodies that might do the trick, the researchers used a panel of seven antibodies trained against different parts of the tau protein, or against specific phospho-residues, to deplete tau from the brain extracts, then tested the remaining seeding activity. They found that some antibodies quashed seeding more effectively than others, and that there was significant variability between samples. Overall, antibodies such as AT8 and PHF1, which bind to pathological forms of tau, inhibited seeding most consistently across samples.

At first glance, the heterogeneity in tau’s seeding capacity and biochemical forms across AD brains may seem at odds with cryo-electron microscopy studies, which identified two predominant conformations of tau fibrils in people with AD (Jul 2017 news). That’s not the case, Dujardin noted. Soluble tau oligomers—not fibrils—are the source of tau’s biochemical variability in this postmortem study. He said it would be fascinating to examine tau oligomers via cryoEM.

Lashuel noted that the researchers did not analyze insoluble fractions of tau in their assays, biasing them to zero in on soluble species. He suggested that attention be paid to understanding the near-total lack of seeding activity in the "low seeders," who still ultimately developed AD. Perhaps more answers would be found in insoluble fractions, he said.

Why is one person’s tau not like another’s? Dujardin suspects genetic differences that influence cellular processes such as degradation and autophagy, which may selectively degrade certain forms of tau. Though microglial function theoretically influences these pathways, Dujardin noted that inflammatory markers in the 32 brains did not correlate with seeding activity, at least at this end stage of disease. Individual differences in kinase activity could also influence which phosphorylations tau accumulates over a person’s lifetime.

How might researchers leverage the findings to inform a person’s prognosis? Dujardin noted several possibilities. A handful of studies have managed to detect seeding activity in tau derived from CSF, though Dujardin noted that these assays need to become more sensitive (Takeda et al., 2016). In lieu of directly measuring seeding activity, perhaps quantification of tau oligomers, and/or specific phospho-tau species associated with seeding activity, could serve the same purpose.—Jessica Shugart


  1. This is a well-designed study underlining once more the importance of soluble tau oligomeric assemblies over long tau filaments in Alzheimer’s disease, as well as heterogeneity in tau oligomers. The findings of this study hold significance as varied tau seeding correlates with the clinical measures of the aggressiveness of the disease.

    Manipulation of in vitro conditions can result in indefinite numbers of aggregate species/conformers/strains. This is true for full-length proteins and fragments, so many of the atomic structures of amyloid fibrils were solved. In vivo, extra levels of complexities are present, where genetics, environmental factors, and other amyloids play crucial roles in oligomer formation (Gerson et al., 2016). 

    The present study clearly demonstrates that soluble oligomers from different AD patients seed differently in the tau biosensor cells and MAPT primary neurons, though it did not assess regional vulnerability for soluble tau oligomers in vivo. The authors also showed that the tau oligomers are the most potent seeding species over tau fibrils/filaments, which was recently shown by us and other laboratories.

    It is noteworthy that these oligomers may possess a common mechanism of toxicity and perhaps common structural features. In a study by Marc Diamond’s group, a homogeneity was observed in the insoluble tau aggregates from six AD cases studied (Sanders et al., 2014). Though we still do not know much about soluble tau aggregates, when our group isolated tau oligomers from 15 AD cases, we saw little heterogenicity based on in vitro analyses, but this work is still ongoing. 

    This is not a critique of the nice work presented in this paper. I think even if different individuals with “typical” AD have distinct biochemical features of tau, this does not mean that each individual has a distinct, biologically active tau strain. If I can take a guess, I would say there are no more than six to 12 distinct biologically active 3R/4R tau strains in AD. 

    It is also important to keep in mind that although multiple bioactive conformers/strains can coexist, these species are competing among each other. If this is the case for tau, then probably the most effective oligomer seeds will prevail and dominate other species. They should be considered the most biologically relevant and will reflect the AD clinical representation associated with it. Unfortunately, this is still not investigated for tau and other amyloids.

    Fortunately, the NIA is currently actively supporting studies on tau and amyloid polymorphism through specific RFAs. This is very important for the advancement of the field, as it will help us understand disease mechanisms, progression, and develop therapeutics and diagnostics.


    . Potential mechanisms and implications for the formation of tau oligomeric strains. Crit Rev Biochem Mol Biol. 2016 Nov/Dec;51(6):482-496. Epub 2016 Sep 21 PubMed.

    . Distinct tau prion strains propagate in cells and mice and define different tauopathies. Neuron. 2014 Jun 18;82(6):1271-88. Epub 2014 May 22 PubMed.

  2. There is an entire field of exploration to understand the nature of pathological tau species in AD amongst monomeric, oligomeric, and fibrillary and modified tau species. The study by Dujardin et al. provides new insight to posit that seed-competent tau species are soluble and oligomeric more than monomeric, with a specific pattern of phosphorylation at specific sites including the AT8 epitope. Dujardin’s results also suggest some tau-phosphorylated species could be protective against seeding activity.

    Significantly, the authors clarified that phosphorylation at the pT181 and pT217 sites found in sedimented brain extract does not correlate with tau seeding activity, age of disease onset, or rate of disease clinical progression. These sites are well-established markers to diagnose AD in cerebrospinal fluid (Mar 2020 news; Mar 2020 news) and are highly promising in plasma (Dec 2019 conference news). The current study demonstrates that pT181 and pT217, hyperphosphorylated in CSF during early disease stages, would not necessarily be pathological or toxic tau species. This result may be fundamental for refining our interpretation of tau biomarker changes.

    The authors focused their investigation on the main tau-phosphorylated sites found in brain. Future studies characterizing species prone to exacerbating tau seeding may also investigate other numerous tau post-translational modifications, including other phosphorylated sites, tau truncation, ubiquitination, acetylation, and methylation. For example, microtubule binding regions (MTBR) are necessary for the pathological aggregation of tau (Kadavath et al., 2015), which is initiated by conversion of tau monomers containing specific motifs in the MTBR from an inert to a seed-competent form (Mirbaha et al., 2018). This region is also the main component of tau aggregates (Fitzpatrick et al., 2017). In the future, it would be really interesting to explore potential enrichment of the MTBR domain in seeding competent tau using dedicated assays.

    Notably, the authors also demonstrated that reduction of tau seeding can be obtained by some of the tested antibodies and could be subject-dependent. The AT8 antibody seems promising. It is worth noting that AT8-related phosphorylations have been described to promote tau self-aggregation in vitro (Despres et al., 2017). Future identification of pathological tau species in brain extracellular spaces or CSF may support the consideration of therapies using corresponding antibodies. There may be heterogeneity in tau-seeding properties, conformation, and post-translational modifications among AD individuals. Overall, this report strongly supports the mitigation of seed-competent species as a promising approach for therapy. 


    . Tau stabilizes microtubules by binding at the interface between tubulin heterodimers. Proc Natl Acad Sci U S A. 2015 Jun 16;112(24):7501-6. Epub 2015 Jun 1 PubMed.

    . Inert and seed-competent tau monomers suggest structural origins of aggregation. Elife. 2018 Jul 10;7 PubMed.

    . Cryo-EM structures of tau filaments from Alzheimer's disease. Nature. 2017 Jul 13;547(7662):185-190. Epub 2017 Jul 5 PubMed.

  3. These are interesting and rather compelling findings. It is nice to see that antibody-mediated targeting of phospho-tau epitopes, as we have advocated, was most effective under these conditions.

    It is notable that the phospho-epitopes identified by mass spectroscopy to correlate positively or negatively with seeding do not relate well with efficacy of antibodies targeting these or closely related epitopes in preventing seeding. This seems to indirectly support our prior findings that antibodies against the same epitope can differ greatly in their efficacy (Congdon et al., 2016Congdon et al., 2019). 

    I concur with the authors’ view that the data suggest benefits of personalized tau treatment, for which brain imaging with tau antibody fragments should provide valuable information on the accessible tau epitope profile of each individual (Krishnaswamy et al., 2014). 


    . Affinity of Tau antibodies for solubilized pathological Tau species but not their immunogen or insoluble Tau aggregates predicts in vivo and ex vivo efficacy. Mol Neurodegener. 2016 Aug 30;11(1):62. PubMed.

    . Tau antibody chimerization alters its charge and binding, thereby reducing its cellular uptake and efficacy. EBioMedicine. 2019 Apr;42:157-173. Epub 2019 Mar 22 PubMed.

    . Antibody-derived in vivo imaging of tau pathology. J Neurosci. 2014 Dec 10;34(50):16835-50. PubMed.

  4. In the early days of research on the role of Tau protein in AD, Brad Hyman and colleagues showed that the distribution of neurofibrillary tangles in the brain was a better predictor of the progression of AD than senile plaques (Arriagada et al., 1992), consistent with the AD staging scheme of Braak & Braak (Braak and Braak, 1991). This was an important milestone in the evolving saga of “Tauists vs Baptists.”

    Now, about 30 years later, Hyman and colleagues put the Tau-AD relationship on a much more refined and quantified basis (Dujardin et al., 2020). Their data are based on a rigorously characterized set of AD patient brains and Tau tangle preparations derived from them postmortem. They were analyzed by multiple biochemical and functional assays such as Tau content, state of aggregation (monomers, oligomers, aggregates), state of phosphorylation, reactions with AD-diagnostic antibodies, and ability to induce local Tau inclusions in sensor cells developed by Marc Diamond and coworkers’ “seeding assay,” reporting on an abnormal cellular distribution of a GFP-labeled Tau repeat domain expressed in HEK cells (Holmes et al., 2014). 

    One question in the study was how the variations in AD on a patient level, e.g., age of onset, duration, rate of progression, etc., correlated with the Tau-containing tangle preparations. The result was that 25 percent of patient variability could be explained by variations in Tau-dependent parameters. The variations were seen on the level of neurons (as expected, as Tau is a neuronal protein), but also on the level of astrocytes (also expected, considering the interactions between these cell types, inflammatory processes, Aβ-dependent effects, and more). In particular, the results showed that the extent of the “seeding reaction,” measured both in terms of rate and final level of inclusions in sensor cells, correlated well with the rate of disease progression in patients. This confirms that this widely used assay is a valuable tool in AD research.

    One caveat in the interpretation of the seeding reaction, common to this paper and many others in the field, is that the appearance of local accumulations of TauRD-GFP is assumed to arise from the templated self-assembly of AD-like Tau filaments. This does not affect the conclusions of this paper, but it does affect the concepts of how “Tau pathology” arises and spreads in the brain. The idea is that Tau molecules in some state of conformation, termed “misfolded,” and aggregation, e.g., oligomeric, are transferred from one neuron to the next and convert the Tau in recipient cells into AD-like fibers. However, the inclusions of TauRD-XFP in sensor cells are observed by fluorescence transfer (FRET), which only reports on a loose vicinity of molecules, up to ~10nm. The large size of the attached GFP (~3x4nm) is a strong steric hindrance for amyloid-like aggregation, which requires a 0.47 nm distance between molecules (Kaniyappan et al., 2020). This steric blocking effect is common with different amyloid-forming proteins whose aggregation is prevented by labeling with GFP or other protein tags.

    Therefore the inclusions in sensor cells must arise from mechanisms distinct from Tau filament formation, for example, as components of stress granules, or possibly as liquid dense clusters formed by phase separation, neither of which require fibers with β-structure (Vanderweyde et al., 2016; Wegmann et al., 2018). Consistent with this, the “seeding reaction” works efficiently only with preparations from AD brain which are rich in non-Tau components, and the reaction can be induced by non-Tau triggers such as cytokines (Gorlovoy et al., 2009). 

    The situation is reminiscent of the cytoplasmic inclusions of FUS-GFP. They can be triggered by stress, e.g., arsenite, leading to stress granules (Marrone et al., 2018). The implications are (a) that factors other than Tau, but closely associated with it, may act as triggers of Tau pathology (e.g. RNA, Fichou et al., 2018), and (b) that spreading of Tau protein between cells may not be necessary for causing Tau pathology.


    . Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer's disease. Neurology. 1992 Mar;42(3 Pt 1):631-9. PubMed.

    . Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991;82(4):239-59. PubMed.

    . Tau molecular diversity contributes to clinical heterogeneity in Alzheimer's disease. Nat Med. 2020 Aug;26(8):1256-1263. Epub 2020 Jun 22 PubMed. Correction.

    . Cofactors are essential constituents of stable and seeding-active tau fibrils. Proc Natl Acad Sci U S A. 2018 Dec 26;115(52):13234-13239. Epub 2018 Dec 11 PubMed.

    . Accumulation of tau induced in neurites by microglial proinflammatory mediators. FASEB J. 2009 Aug;23(8):2502-13. PubMed.

    . Proteopathic tau seeding predicts tauopathy in vivo. Proc Natl Acad Sci U S A. 2014 Oct 14;111(41):E4376-85. Epub 2014 Sep 26 PubMed.

    . FRET-based tau seeding assay does not represent prion-like templated assembly of tau fibers. bioRxiv March 25, 2020. BioRxiv.

    . Isogenic FUS-eGFP iPSC Reporter Lines Enable Quantification of FUS Stress Granule Pathology that Is Rescued by Drugs Inducing Autophagy. Stem Cell Reports. 2018 Feb 13;10(2):375-389. Epub 2018 Jan 18 PubMed.

    . Interaction of tau with the RNA-Binding Protein TIA1 Regulates tau Pathophysiology and Toxicity. Cell Rep. 2016 May 17;15(7):1455-1466. Epub 2016 May 6 PubMed.

    . Tau protein liquid-liquid phase separation can initiate tau aggregation. EMBO J. 2018 Apr 3;37(7) Epub 2018 Feb 22 PubMed.

  5. Dujardin et al. found a correlation between the rate of progression of Alzheimer’s disease and the seeding activity of tau extracted from postmortem samples of frontal cortex (Brodmann areas 8-9) in 32 cases. The seeding activity was evaluated in a “biosensor”—a specially developed cell line in which tau seeds induce the formation of fluorescent aggregates, emitting a FRET signal. Live imaging of the formation of aggregates showed an ascending phase and a plateau. The plateau is considered dependent on the number of seeds added to the culture.

    The variability of the seeding activity measured with the biosensor cell line was controlled by a second experiment using primary culture of neurons from transgenic mice expressing different human tau isoforms. Injection of brain extracts to P301S transgenic mice showed a higher number of AT8 positive neurons in the animals that received the extracts that had a higher seeding efficiency in vitro.

    Tau burden in the brains of the patients was also correlated with tau seeding activity. The seeding activity was concentrated in a high molecular weight fraction after size-exclusion chromatography; it was associated with the presence of oligomers (HT7/HT7 assay recognizing multimeric tau) and of specific phosphorylated tau sites, detected by both mass spectrometry and immunoassays.

    This remarkable paper raises several crucial questions. In our view, the most difficult one is the distinction between quality and quantity of tau seeds: Is the rate of progression of Alzheimer’s disease related to various tau “strains” (quality of the seeds), or to a variation in the amount of tau seeds (quantity of the seeds)? In other words, are the seeds structurally different from one case to another, or are they identical but in different concentrations?

    The paper provides evidence that there are indeed more seeds in the cases with a rapid progression. Evidence is still lacking that the seeds are structurally different: The differential sensitivity to proteinase K may partially support that contention, but it is probable that technical progress will be necessary to reach a better understanding of the mechanism—qualitative or quantitative—at work in the heterogeneity of Alzheimer’s disease.

    Several other more technical questions may be raised:

    • The age of the patients is very diverse. Several are young (six below 65 and 13 below 70). Were mutations systematically sought in those young cases?
    • The rate of progression is taken as a reliable characteristic of the disease itself, dependent on tau seeding activity. Many factors could, however, play a role. Extended figure 7 shows that the highest seeding activities were found in the youngest cases and were associated with the highest tau burdens (extended figure 6). An inverse causation is possible: In this interpretation, it would be wrong to consider that the progression is faster because there is more seeding activity. The correct interpretation could be that more seeding activity is detected because, the progression being rapid, the samples may contain more seeds that tend to disappear over time during the course of the disease. The progression of the disease, in this view, could be the independent variable, explaining the seeding activity (Seeding activity = a x disease duration + b) rather than the contrary as suggested in the paper (Duration of the disease = a x seeding activity + b).
    • The positive correlation between the clinical symptoms and the various indices of tau seeding activities is often driven by a few cases with slow progression and low indices (i.e., cases at an older age). See for instance figure 4 c, or extended figure 9: no correlation is left if the few cases with low burden and very slow course were taken out. In this respect, the distinction between “hippocampal sparing” cases and “limbic predominant” cases (Murray et al., 2011) may be interesting.

    Those specific points do not call into question the great merits of the paper, which shows that the variability in the seeding activity measured in the adequate biosensor has a real biological significance and that the tau species that seed the pathology are phosphorylated tau oligomers.

    We take into account the recent paper by Mandelkow’s group that showed that the biosensor assay does not reproduce the in vivo situation (Kaniyappann et al., 2020). But we also underline, as do Dujardin et al., that paired helical filaments are probably not the “seeds,” and that the variation in the seeding activity shown by the biosensor probably has a biological significance. In a recent investigation, we were surprised to find out that a tau seeding activity could be detected in the absence of paired helical filaments, was associated with phosphorylated species and not causally linked to amyloid (Thierry et al., 2020). The paper by Dujardin et al. opens the ways to new strategies to elucidate tau pathology and its progression.


    . Neuropathologically defined subtypes of Alzheimer's disease with distinct clinical characteristics: a retrospective study. Lancet Neurol. 2011 Sep;10(9):785-96. PubMed.

    . FRET-based tau seeding assay does not represent prion-like templated assembly of tau fibers. bioRxiv March 25, 2020. BioRxiv.

    . Human subiculo-fornico-mamillary system in Alzheimer's disease: Tau seeding by the pillar of the fornix. Acta Neuropathol. 2020 Mar;139(3):443-461. Epub 2019 Dec 10 PubMed.

  6. It is well established that the severity of cognitive impairment in Alzheimer’s disease correlates with the number of tau inclusions. Brad Hyman played an important part in helping to establish this concept (Hyman et al., 1984; Arriagada et al., 1992; de Calignon et al., 2012). It has also been known for some time that the rate of cognitive decline and the duration of disease can vary substantially between patients. Dujardin, Hyman, and colleagues now put forward the intriguing hypothesis that this heterogeneity may be explained, at least in part, by the presence of different tau species of unknown structure, rather than by variable amounts of the tau filaments whose structures we determined from frontal cortices of patients with a neuropathologically confirmed diagnosis of Alzheimer’s disease (Fitzpatrick et al., 2017Falcon et al., 2018). 

    Dujardin and colleagues homogenized frontal cortices from 32 patients who had died with neuropathologically confirmed diagnoses of AD, spun the samples at 10,000 g for 10 min, and used the supernatants in several in vitro seeding assays. Despite the addition of a constant amount of total tau, seeding activity varied by as much as 10-fold between cases, giving rise to the grouping into high, medium, and low seeders. Dujardin and colleagues conclude that seeding from Alzheimer’s disease brain was caused by a soluble, high-molecular weight species of tau, in agreement with earlier reports from the same group (Takeda et al., 2015, 2016). 

    Next, they injected assemblies from high, medium, and low seeders into mice transgenic for human P301S tau. Assemblies from high seeders gave rise to more AT8-positive cells than those from medium and low seeders. For these experiments, brain homogenates were spun at 150,000 g for 30 min and the resuspended pellet fractions injected. These findings are reminiscent of those reported earlier in another transgenic mouse line injected with tau seeds (Clavaguera et al., 2009). Seeding activity resided almost entirely in the pellet fraction following an ultracentrifuge spin (100,000 g for 20 min) of the brain homogenates, leading us to conclude that most seeding activity was present in the insoluble fraction.

    What is soluble to some, is insoluble to others. We define soluble tau as that which stays in solution following a high-speed spin (100,000 g or higher). The pellet contains insoluble tau. But it may be preferable to refer to molecular tau species instead. Dujardin and colleagues describe a strong correlation between seeding activity and the amount of high-molecular weight tau. This conclusion is based on their use of an HT7/HT7 AlphaLISA assay, which gives a positive signal when at least two tau molecules bind to each other. They claim that this assay quantifies oligomeric, rather than all multimeric, tau. We previously used sucrose gradient fractionation and HEK293T cells inducibly expressing 1N4R P301S tau to show that short tau filaments were the most seed-competent species in brain extracts from mice transgenic for human P301S tau (Jackson et al., 2016). Other high-molecular weight tau species (>10mers) were also seed-competent, but low-molecular-weight and monomeric tau were unable to seed assembly. When injected into mice transgenic for human P301S tau, the fractions that seeded assembly in HEK293T cells also promoted tau assembly.

    These findings raise the question of what happened to tau filaments in the study by Dujardin and colleagues. Neuropathology had shown that the frontal cortices from their cases of Alzheimer’s disease carried many filamentous tau inclusions (Braak stages V/VI). Although insoluble tangle fragments may pellet following a 10 min spin at 10,000 g, this is unlikely to be the case for dispersed filaments. In fact, when we extract tau filaments from the brains of patients with Alzheimer’s disease, we routinely use supernatants of brain homogenates spun at 20,000 g for 30 min as the source of dispersed filaments.

    Hopefully, future experiments will tell us what the relative contributions of high-molecular weight species (including oligomers and filaments) are to tau spreading in the brains of patients with Alzheimer’s disease, and if these species have shared or different conformations.


    . Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer's disease. Neurology. 1992 Mar;42(3 Pt 1):631-9. PubMed.

    . Transmission and spreading of tauopathy in transgenic mouse brain. Nat Cell Biol. 2009 Jul;11(7):909-13. PubMed.

    . Propagation of tau pathology in a model of early Alzheimer's disease. Neuron. 2012 Feb 23;73(4):685-97. PubMed.

    . Tau filaments from multiple cases of sporadic and inherited Alzheimer's disease adopt a common fold. Acta Neuropathol. 2018 Nov;136(5):699-708. Epub 2018 Oct 1 PubMed.

    . Cryo-EM structures of tau filaments from Alzheimer's disease. Nature. 2017 Jul 13;547(7662):185-190. Epub 2017 Jul 5 PubMed.

    . Alzheimer's disease: cell-specific pathology isolates the hippocampal formation. Science. 1984 Sep 14;225(4667):1168-70. PubMed.

    . Short Fibrils Constitute the Major Species of Seed-Competent Tau in the Brains of Mice Transgenic for Human P301S Tau. J Neurosci. 2016 Jan 20;36(3):762-72. PubMed.

    . Neuronal uptake and propagation of a rare phosphorylated high-molecular-weight tau derived from Alzheimer's disease brain. Nat Commun. 2015 Oct 13;6:8490. PubMed.

    . Seed-competent high-molecular-weight tau species accumulates in the cerebrospinal fluid of Alzheimer's disease mouse model and human patients. Ann Neurol. 2016 Sep;80(3):355-67. Epub 2016 Aug 3 PubMed.

  7. This is a detailed human tissue study that reports on the heterogeneity of pathogenic tau species in spite of near-identical histological scoring of NFTs. While done only in 32 subjects with “typical” AD, it’s already clear that the biochemical and biological activity of pathogenic tau has significant inter-subject variation—imagine how convoluted the data becomes when the sample size is increased.

    I am glad to see that several key findings in this paper are in agreement with our study published May 2019 in Science Translational Medicine (Aoyagi, et al., 2019). Using a panel of analogous HEK cell-based bioassays, we probed for the prion-like activity of tau, α-synuclein, and Aβ in 114 patient samples from sporadic and familial forms of AD and CAA, as well as several cases of FLTD-tau and MSA used as controls for propagation specificity. Moreover, we performed ELISA for several species of tau and Aβ in soluble and insoluble brain fractions.

    Development of a new cellular bioassay for Aβ has enabled parallel quantification of Aβ- and tau-prion, aka “seeding” activity, providing the first direct quantitative comparison of the active propagating species, rather than inert protein deposits. Our data show that AD patients (Braak V/VI) with greatest longevity have lower amounts of both Aβ- and tau-prion activity at the time of death than patients who die at younger ages from AD-related neurological dysfunction.

    Long-lived individuals with low amounts of tau-prion activity correlated with the levels of phosphorylated tau and were not carriers of the APOEε4 allele. Surprisingly, the longevity-dependent decrease in tau-prion activity occurred in spite of increasing levels of insoluble total tau. When corrected for the abundance of insoluble total tau, the tau-prion activity decreased exponentially with respect to the age of death. Moreover, the half-time was approximately one decade, and this correlated with the abundance of phosphorylated tau. In other words, early onset AD patients who died in their 40s had approximately 32 times higher “specific activity” of tau-prions compared to late-onset AD patients who died in their 80s and 90s.

    In summary, the findings from our group and Dujardin et al. strongly argue that:

    1. Histologic and biochemical measurements of protein abundance alone can be deceptive and provide an incomplete picture of disease pathogenesis.
    2. Replication-competent species of tau and Aβ may track better with progression and become more relevant targets for therapy.

    Lastly, because the debate on the relevance of HEK cell assays with YFP-fusion proteins has infiltrated this commentary, it seems prudent to say something brief here. As others have expressed, given the large size of the reporter protein and type of cell, there was always concern about the fidelity of transmission—does the output match the input? However, if considering these cells simply as a “biosensor,” then detecting the presence of replication-competent tau, or whatever aberrant protein measured, is still informative even if imperfect.

    Aoyagi, Condello, et al. report that the tau, α-synuclein, and Aβ bioassays exhibit homotypic propagation, similar to earlier studies in the Diamond and Prusiner labs. Using brain extracts from transgenic mice that produce only human tau, synuclein, or Aβ aggregates, we show that:

    1. tau-YFP cells only respond to brain-derived tau-prions;
    2. synuclein-YFP cells only respond to brain-derived synuclein-prions; and
    3. YFP-Aβ cells only respond to brain-derived Aβ-prions.

    Importantly, using human brain samples, we show that synuclein-prions in MSA induce a response in a synuclein bioassay but not the other cells; similarly, tau-prions in FTLD-tau cases induce a response in tau bioassay but not the other cells. Given the omnipresence of neuroinflammation in these rodent models and neurodegenerative disorders, it seems unlikely that generic cytokines or other inflammatory factors alone could prompt the formation of YFP-positive puncta and yield a false positive. Having said that, propagation and spreading do not happen in a vacuum. The brain is the most complex organ, and it is expected and already becoming known which genetic and cellular factors influence these processes in vivo.


    . Aβ and tau prion-like activities decline with longevity in the Alzheimer's disease human brain. Sci Transl Med. 2019 May 1;11(490) PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Connectivity, Not Proximity, Predicts Tau Spread
  2. Longitudinal Tau PET Links Aβ to Subsequent Rise in Cortical Tau
  3. Serial PET Nails It: Preclinical AD Means Amyloid, Tau, then Cognitive Decline
  4. Cellular Biosensor Detects Tau Seeds Long Before They Sprout Pathology
  5. Widely Used Tau Seeding Assay Challenged
  6. A Phospho-Tau Plasma Assay for Alzheimer’s?
  7. 217—The Best Phospho-Tau Marker for Alzheimer’s?
  8. Tau Filaments from the Alzheimer’s Brain Revealed at Atomic Resolution

Paper Citations

  1. . Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991;82(4):239-59. PubMed.
  2. . High degree of heterogeneity in Alzheimer's disease progression patterns. PLoS Comput Biol. 2011 Nov;7(11):e1002251. PubMed.
  3. . Seed-competent high-molecular-weight tau species accumulates in the cerebrospinal fluid of Alzheimer's disease mouse model and human patients. Ann Neurol. 2016 Sep;80(3):355-67. Epub 2016 Aug 3 PubMed.

Further Reading


  1. . Cerebrospinal fluid from Alzheimer's disease patients promotes tau aggregation in transgenic mice. Acta Neuropathol Commun. 2019 May 7;7(1):72. PubMed.

Primary Papers

  1. . Tau molecular diversity contributes to clinical heterogeneity in Alzheimer's disease. Nat Med. 2020 Aug;26(8):1256-1263. Epub 2020 Jun 22 PubMed. Correction.