A diverse bunch of neurodegenerative diseases, tauopathies are marked by a range of clinical manifestations and of neuropathological hallmarks. According to a study published September 29 in Nature, the defining feature of each type of tauopathy may boil down to a minute detail that requires a cryoelectron microscope to see. That detail is the way each molecule of tau contorts itself within a protofibril. Led by Michel Goedert and Sjors Scheres of the MRC Laboratory of Molecular Biology, Cambridge, England, U.K., the study describes an array of high-resolution structures of tau protofibrils plucked from postmortem brain samples. Tau folds tracked remarkably well with the neuropathological characteristics of each disease. Finally, the researchers used these folds to create a hierarchical classification of different diseases, akin to a phylogenetic tree.

  • New cryo-EM structures for a variety of 4R-tauopathies show distinct folds of tau.
  • In globular glial tauopathy and PSP, each tau molecule folds into three layers.
  • For several other tauopathies, tau folds into four layers.
  • These distinct folds enable a hierarchical classification of tauopathies.

“[These studies] present new evidence for a link between the molecular conformation of the tau protein and the resulting clinicopathologic phenotype,” commented Lary Walker of Emory University, Atlanta. “These seemingly disparate disorders are united by a common thread, i.e., that the key to a great many neurodegenerative disorders lies in the problem of altered protein structure.”

The study is the culmination of years of structure-solving by Scheres and Goedert. Wielding the tool of cryo-EM, they first took on one tauopathy at a time. In 2017, they dazzled the AD field when they described the back-to-back C-shaped structures that stacked in paired helical and straight filaments of that disease (Jul 2017 news). Then, they unveiled the wildly different, J-shaped structure of tau in filaments from a person with Pick’s disease (Aug 2018 news). Next, they zeroed in on tau filaments from people with chronic traumatic encephalopathy, uncovering a C-shaped fold similar to that of AD, but with a mysterious molecule nestled within (Mar 2019 news). Most recently, they reported yet another flavor of fold—a quadruple-layered contortion found in three people with corticobasal degeneration (Feb 2020 news). 

Folds of a Feather Flock Together. Dendrogram of a proposed structure-based classification of tauopathies. Colors denote different microtubule-binding domains R1-R4, flat arrows denote β-strands. Non-protein entities are in black. [Courtesy of Shi et al., Nature 2020.]

The new paper is a grand finale of sorts, describing and comparing structures from several tauopathies at once. It connects structures and the neuropathological and clinical syndromes with which they associate. Most of the new structures are derived from individuals with different four-repeat tauopathies. In these diseases, tau filaments are made up of tau proteins that repeat tau's microtubule-binding region four times. This is in contrast to Pick’s disease, in which tau filaments contain only three repeats, and mixed tauopathies such as AD and CTE, in which both 3R- and 4R-tau weave into fibrils.

Co-first authors Yang Shi and Wenjuan Zhang and colleagues resolve the structures of tau filaments from the brains of people with the 4R tauopathies progressive supranuclear palsy (PSP), globular glial tauopathy (GGT), agyrophillic grain disease (AGD), aging-related tau astrogliopathy (ARTAG), and from people with MAPT intron-10 mutations +3 and +16, which cause a form of frontotemporal degeneration (FTD). The scientists also reported structures of two mixed tauopathies—familial British dementia (FBD) and familial Danish dementia (FDD).

First, they took on PSP-tau. PSP is a sporadic 4R tauopathy that manifests in typical and atypical clinical forms. People with the typical form, also called Richardson’s syndrome, suffer loss of balance, lack of eye coordination, trouble swallowing, and a range of other cognitive and behavioral symptoms. At 2.7Å resolution, the researchers solved the structures of tau filaments from three people with PSP-RS (see image below). They also solved the structure of tau from two people with an atypical, frontal presentation (PSP-F), one with predominant parkinsonism (PSP-P), and one with corticobasal syndrome (PSP-CBS).

Hello Tau. Cryo-EM map of a tau fibril taken from a PSP patient. [Courtesy of Shi et al., Nature, 2021.]

Except for a distinct structure in one PSP-F patient, Shi identified the exact same tau fold in all the people with PSP. Stacked in rungs into a single protofilament, the PSP-fold comprised three layers involving the three microtubule-binding domains. R3 formed the central layer, while R2 and R4 packed on either side, sandwiching R3 along with some additional, unknown molecules (see image below).

Next, the scientists examined tau fibrils extracted from brain samples of three people with GGT, another sporadic 4R tauopathy. The same “GGT-fold” was common to all three. This fold was similar to the PSP-fold in that it comprised a three-layered sandwich of R2-R4 domains. Beyond this, there were distinguishing characteristics at every turn. Each bend comprised β-sheets of a different number and length compared to those in the PSP-fold, and the C-terminus of GGT-tau pointed in the opposite direction of its counterpart in PSP. While all tau molecules from the GGT samples had the same fold, they wound into three different fibril configurations. Type 1 protofilaments, akin to PSP filaments, were made up of a single stack of taus. Type 2 packed two protofilaments symmetrically, and type 3 packed two protofilaments asymmetrically (see image below).

Interestingly, tau from one of the two PSP-F volunteers folded into a shape that resembled aspects of the PSP-fold and the GGT-fold. The researchers dubbed this amalgam the “GPT-fold.” It also came in both single and double protofilament versions—type 1 and type 2, respectively.

Triple Threat. Three-layered tau folds, arranged in single protofilaments or in different configurations of doubles, were found in PSP and GGT brains. In one PSP patient, tau bent into a hybrid fold called GPT. [Courtesy of Shi et al., Nature, 2021.]

Tau extracted from the brains of two people who had died with AGD took a markedly different fold than the PSP, GGT, or GPT ones. It formed four layers instead of three. Akin to the previously identified CBD fold, the quadruple-layered AGD fold involved R2-R4, with the C-terminus folding back onto R2 to create the fourth layer (see image below). AGD filaments came in three configurations with a common protofilament core. Type 1 was a single protofibril; type 2 stacked two protofibrils with C2 screw symmetry, and type 3 was different from the other two, but its structure was not resolvable enough to decipher the details.

Fourfold. The “AGD-fold” comprised four layers, with the C-terminus (red) snuggling up to R2 (blue), which folded against R3 (green). R4 (yellow) wrapped around R3 and spooned a turn formed by R2-R3. The fold came in two filament structures—one made of a single protofilament (left), and one that stacked pairs of protofilaments (right). [Courtesy of Shi et al., Nature, 2021.]

The AGD-fold made another appearance in samples from people with ARTAG. This jibes with the existence of granular or fuzzy astrocytes in gray matter in both diseases. Tau wound into the AGD fold in filaments from three people with MAPT intron-10 mutations +3 or +16, as well. Notably, argyrophilic grains have been spotted in people with these MAPT mutations, suggesting a neuropathological overlap between AGD and this type of FTD.

Thus, the 4R tauopathies fell into two categories based on the number of layers within their tau folds. PSP, GGT, and GPT all had taus with three-layered folds, while AGD, ARTAG, and the MAPT mutations all shared the same four-layered fold. CBD also falls into the quadruple-layer category.

Finally, Shi, Zhang et al. characterized tau’s shape in people with two inherited tauopathies: FBD and FDD. These mixed 3R+4R tauopathies are caused by mutations in the integral membrane protein 2B gene (ITM2B), which leads to the production of amyloidogenic peptides, akin to the role of Aβ in AD. These then spark tauopathy. The scientists found paired helical filaments in which tau folded into the characteristic back-to-back C-shaped structures found in AD. One of the patients also had straight helical filaments identical to those found in AD, as well as CTE-type filaments, possibly due to a prior head trauma.

Pulling together all the tau structures identified in this paper and its predecessors, the researchers propose a hierarchical classification of tauopathies based on their tau folds (see dendrogram above). Mixed 3R+4R tauopathies formed one branch, with AD, FBD, FDD, and primary age-related tauopathy sharing the same tau fold, and CTE-tau forming a similar, but distinct fold. Pick’s disease—a pure 3R tauopathy—remained on its own with its J-shaped tau fold. The 4R tauopathies split into two branches based on the number of layers in their folds.

This effort organizes the increasingly complex taxonomy of tauopathies, commented James Rowe of Cambridge University. “The new hierarchical dendrogram that classifies and subtypes the family of tauopathies according to microstructure folds is not only commended for its inclusion of both common and rare tauopathies, but also for providing a principled way to predict and confirm new tauopathies. It will help the design and stratification of new anti-tau therapeutics,” Rowe wrote.

Importantly, the folds aligned with neuropathologically confirmed diagnoses. They also cemented distinctions between disorders that previously had been lumped together based on clinical symptoms. For example, PSP and CBD have been grouped together clinically, despite their distinct patterns of neuropathology. That their tau folds belong to two different classes—three- versus four-layered folds—supports the idea that they are distinct disorders. Although PSP and GGT both had three-layered tau folds, the subtler differences between these folds supports the idea that they are distinct disorders, as well.

Finally, the outlier—the hybrid “GPT-tau” fold found in one PSP-F case. It suggests that this person may have had a disorder other than PSP. Based on the distribution of neuropathology in this case, the researchers suggest calling the disease “limbic inclusion body 4R tauopathy (LNT).”

The new classification scheme has significant implications for biomarker and drug development, noted Gil Rabinovici of the University of California, San Francisco. “For example, that aggregates in PSP and CBD, previously considered collectively as ‘4R tauopathies,’ have different folding structures suggests that the fibrils in these disorders may interact very differently with small molecules that are being developed as PET tracers or therapeutics,” he wrote. “The heterogeneity of tauopathies may be even greater than we previously imagined, which is a bit daunting but also incredibly informative for future work.”

What makes tau bend into distinct conformations in different diseases? This is the ultimate question that has yet to be answered, Scheres said. Among explanations proffered for tau folds are brain regions or cell types in which tau first misfolds, and different triggers that might initiate the process—such as amyloid, head trauma, or mutations. While mysteries remain, one thing seems clear to Scheres: “It would be hard to imagine that tau aggregates are just a waste product of homeostasis gone wrong,” he said. “That is an argument that is hard to hold in light of these results.”

This structure-based classification of tauopathies aligns well with categorization based on seeding, commented Marc Diamond of the University of Texas Southwestern Medical Center in Dallas. Diamond said the seeding characteristics of tau monomers extracted from the brains of people with different tauopathies support the importance of tau’s folds in dictating disease type (Sharma et al., 2018). 

Along those lines, Mathias Jucker of the German Center for Neurodegenerative Diseases in Tübingen cautioned that tau filaments pulled from postmortem brain samples may be distinct from those that sparked the disease years earlier. “We should remember that these are end-stage structures, and do not necessarily tell us something about the bioactive conformations,” he wrote. “The most active Aβ and tau seeding activity is at the early stage, not at the end stage of disease.”—Jessica Shugart


  1. Tauopathy is present in at least 25 neurodegenerative diseases (Spillantini and Goedert, 2013), and thus it is one of the most common proteopathies to afflict the nervous system. Postmortem investigations of tauopathies have documented the diversity of structure, cellular vulnerability, anatomical distribution, and chemical makeup of the associated lesions. Recent analyses have delved ever more deeply into the molecular underpinnings of disease variability. These cryo-electron microscopic studies present new evidence for a link between the molecular conformation of the tau protein and the resulting clinicopathologic phenotype. The results highlight the diversity of tau structures, while also showing that tauopathies can be hierarchically classified based on the folds adopted by the protein.

    Some thoughts on this intriguing investigation:

    1. The findings reinforce the conclusion that, although tauopathy is a frequent neurodegenerative phenomenon, it is by no means a unitary disorder clinically, histopathologically, or molecularly. Some causes of variation are coming to light, but many uncertainties remain. Among these, the differential involvement of glial cells and neurons in these various tauopathies is a critical open question.
    2. Exactly how Aβ-proteopathy leads to tauopathy in AD remains unknown. It is interesting that, in familial British dementia, familial Danish dementia, and prion diseases with PrP-amyloid, the tau fold resembles that in AD. For example, the amyloid in familial British dementia consists of a protein that results from a defective stop codon; translation of the elongated open reading frame, and processing of the precursor generates a protein segment with no known function, and yet it, like misfolded Aβ, is highly amyloidogenic (Vidal et al., 1999). Hence, amyloid deposition (“amyloid” in the true, generic sense; Walker, 2020) yields a particular pathologic strain of tau. What common feature of amyloid is responsible for the “AD fold” of tau? Are oligomers involved (Cantlon et al., 2015)? It is also noteworthy that amyloid per se is not essential for driving this particular fold, as it is found in primary age-related tauopathy (PART).
    3. While the different tau folds in different diseases suggest the emergence and spread of disease-specific seeds within the brain, this hypothesis should be tested in experimental animals, the caveat being that truly human-like tauopathy remains elusive in genetically modified rodents. As the authors suggest, cryo-EM studies of experimental animals could inform the translational utility of these models.
    4. This investigation includes a respectable number of subjects for such technically challenging experiments, but future studies should expand the number of both subjects and brain areas examined. A balanced analysis of males and females is required to rule out the possibility of sex differences in the susceptibility to different tau strains. Also, does the stage of disease development influence the molecular phenotype of tau, i.e., do the characteristics of the folds and/or the cell types involved change over time?

    Once again, cryo-EM is showing its value in illuminating the nature of pathogenic proteins. In his seminal presentation in 1906 (Alzheimer, 1907), Alois Alzheimer argued that histopathologic analysis would yield a more precise definition of neurologic diseases than can be gleaned from clinical assessment alone. By defining the molecular configuration of tau polymers derived from multiple tauopathies, Shi and colleagues take Alzheimer's argument a step further: from histopathology to molecular proteopathy.

    To loosely paraphrase Alzheimer, cryo-EM is revealing more molecular subtypes of tauopathy than are currently recognized by our textbooks. That said, these seemingly disparate disorders are united by a common thread, which also happens to define numerous other brain diseases. It is that the key to a great many neurodegenerative disorders lies in the problem of altered protein structure.


    . The Familial British Dementia Mutation Promotes Formation of Neurotoxic Cystine Cross-linked Amyloid Bri (ABri) Oligomers. J Biol Chem. 2015 Jul 3;290(27):16502-16. Epub 2015 May 8 PubMed.

    . Tau pathology and neurodegeneration. Lancet Neurol. 2013 Jun;12(6):609-22. PubMed.

    . A stop-codon mutation in the BRI gene associated with familial British dementia. Nature. 1999 Jun 24;399(6738):776-81. PubMed.

    . Aβ Plaques. Free Neuropathol. 2020;1 Epub 2020 Oct 30 PubMed.

  2. This is a really important paper. It brings clarity and insight into what had begun to seem an ever-more-complex taxonomy of tauopathies since cryo-EM augmented the isoform and clinical classifications. The new hierarchical dendrogram that classifies and subtypes the family of tauopathies according to microstructure folds is not only commended for its inclusion of both common and rare tauopathies, but also for providing a principled way to predict and confirm new tauopathies. It will help the design and stratification of new anti-tau therapeutics.

  3. In my opinion this is an important, paradigm-shifting paper. It reports for the first time the cryo-EM-derived structure of tau filaments in PSP, the prototypical primary tauopathy, and in a variety of other tau-related disorders.

    Building on previous work from Goedert’s and Scheres' labs (in AD, CTE, CBD, and Pick’s), we now have a broad view of similarities and differences in the structure of aggregates across a broad range of tauopathies.

    The novel classification scheme proposed here provides a roadmap for future study of tauopathies that is grounded in structural biology and has significant implications for biomarker and drug development. For example, the fact that aggregates in PSP and CBD, previously considered collectively as “4R tauopathies,” have different folding structures suggests that the fibrils in these disorders may interact very differently with small molecules that are being developed as PET tracers or therapeutics.

    The heterogeneity of tauopathies may be even greater than we previously imagined, which is a bit daunting but also incredibly informative for future work.

  4. This paper represents an important milestone in shifting the definition of sub-entities within the group of tauopathies from syndrome-based, biochemical, or neuropathology-based classification systems to structural biophysical principles.

    This novel approach does not refute the previous classifications, nor does it deprive them of their justification. In a multidimensional classification, the earlier principles remain valid and useful, e.g., for clinical diagnosis and symptomatic therapies, neuropathological diagnosis, and neurobiological studies on the relative contribution of different cell types.

    However, the new biophysical classification seems to provide the stratification that is most upstream within the neurobiology of tauopathies so far. This can open new avenues for the development of diagnostic and therapeutic instruments.

    Whether the suggested tau folds observed in the intracellular aggregates are identical to the spreading or toxic tau samples, i.e., the targets for therapy, remains to be shown.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Tau Filaments from the Alzheimer’s Brain Revealed at Atomic Resolution
  2. Conformers Confirmed: Structure of Pick’s Tau Distinct from AD Tau
  3. Traumatic Tau: Filaments from CTE Share Distinct Structure
  4. CryoEM of CBD Tau Suggests Another Unique Protofibril

Paper Citations

  1. . Tau monomer encodes strains. Elife. 2018 Dec 11;7 PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Structure-based classification of tauopathies. Nature. 2021 Oct;598(7880):359-363. Epub 2021 Sep 29 PubMed.