henhaobio
Sign in

NatureNeurosciencePeer-reviewed journal article

Structure-based Classification of Tauopathies

Yang ShiiD, Wenjuan ZhangiD, Yang YangiD, Alexey G. Murzin, Benjamin FalconiD, Abhay KotechaiD, M. van Beers, Airi TarutaniiD, Fuyuki KametaniiD, Holly J. GarringeriD, Rubén VidaliD, Grace I. HallinaniD, Tammaryn LashleyiD, Yuko SaitoiD, Shigeo MurayamaiD, Mari YoshidaiD, Hidetomo TanakaiD, Akiyoshi KakitaiD, Takeshi IkeuchiiD, Andrew RobinsoniD, David M. A. ManniD, Gábor Géza KovácsiD, Tamás RévésziD, Bernardino Francesco GhettiiD, Masato HasegawaiD, , Sjors H. W. ScheresiD

PDFDOIPreprint
In-depth full-text digest available · Jump to digest ↓

Abstract

Ordered assembly of the tau protein into filaments characterizes multiple neurodegenerative diseases, which are called tauopathies. We previously reported that by electron cryo-microscopy (cryo-EM), tau filament structures from Alzheimers disease (1,2), chronic traumatic encephalopathy (CTE) (3), Picks disease (4) and corticobasal degeneration (CBD) (5) are distinct. Here we show that the structures of tau filaments from typical and atypical progressive supranuclear palsy (PSP), the most common tauopathy after Alzheimers disease, define a previously unknown, three-layered fold. Moreover, the tau filament structures from globular glial tauopathy (GGT, Types I and II) are similar to those from PSP. The tau filament fold of argyrophilic grain disease (AGD) differs from the above and resembles the four-layered CBD fold. The majority of tau filaments from aging-related tau astrogliopathy (ARTAG) also have the AGD fold. Surprisingly, tau protofilament structures from inherited cases with mutations +3/+16 in intron 10 of MAPT, the microtubule-associated protein tau gene, are identical to those from AGD, suggesting that a relative overproduction of four-repeat tau can give rise to the AGD fold. Finally, tau filament structures from cases of familial British dementia (FBD) and familial Danish dementia (FDD) are the same as those from Alzheimers disease and primary age-related tauopathy (PART). These structures provide the basis for a classification of tauopathies that also allows identification of new entities, as we show here for a case diagnosed as PSP, but with abundant spherical 4R tau inclusions in limbic and other brain areas. The structures of the tau fold of this new disease (Limbic-predominant Neuronal inclusion body 4R Tauopathy, LNT) were intermediate between those of GGT and PSP.

In-depth summary

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 1 (28 May 2021) · CC BY 4.0

Cryo-EM structures sort four-repeat tauopathies into three-layered (PSP, GGT, LNT) and four-layered (CBD, AGD, ARTAG, MAPT intron 10 mutations) tau folds, grounding a hierarchical structure-based classification.

Tau isoform content (3R, 4R or both) sorts tauopathies into three broad groups, and some globular glial tauopathy (GGT) cases were once diagnosed as atypical progressive supranuclear palsy (PSP). Extending earlier cryo-EM folds for Alzheimer's disease, Pick's disease and corticobasal degeneration (CBD) to PSP, GGT, argyrophilic grain disease (AGD), aging-related tau astrogliopathy (ARTAG) and inherited dementias yields a fold-based tauopathy classification.

Key results

  1. ShownTypical PSP (Richardson's syndrome) and atypical forms with frontal (one of two cases), parkinsonian or corticobasal-syndrome presentation share one previously undescribed three-layered tau fold, spanning roughly the CBD core residues in a markedly different conformation.

    Maps up to 2.7 Å from three PSP-RS cases plus PSP-F case 1, PSP-P and PSP-CBS; single-protofilament core spans residues 272-381; R2-R4 form a meander turning at PGGG motifs, with R3 bent near-right-angle at G323; R3-R4 cavity lined by K317, K321, K340 holds a large, presumably anionic density. · RESULTS

  2. ShownA four-layered AGD fold, partly matching the CBD fold, forms AGD filaments, most ARTAG filaments and the only filaments seen in MAPT intron 10 +3/+16 carriers; the authors propose relative 4R tau overproduction as a route to this fold.

    Two stage 3 AGD cases (nucleus accumbens), up to 3.4 Å: core residues 273-387 (type 1, single protofilament) or 279-381 (type 2, C2-symmetric pair); residues 293-357 match CBD, core seven residues longer. ARTAG: mostly AGD types 2 and 3. Two +3 and one +16 carriers: AGD fold only. All give 37 kDa C-terminal bands, unlike the 33 kDa band of PSP and GGT. · RESULTS; DISCUSSION; Figure legends

  3. ShownGGT Types I and II share a distinct three-layered fold covering nearly the same residues as PSP but with different PGGG turns and an opposite C-terminal orientation, supporting GGT as a neuropathological entity separate from PSP.

    Frontal cortex maps up to 2.9 Å; core residues 272-379; GGT-I shows three filament types (single protofilament, symmetric pair, asymmetric pair), GGT-II only types 2 and 3; only R2 273-285 and R3 322-330 match PSP; the C-terminal hairpin resembles the Pick fold; R3-R4 cavity larger and more positive (E338-K331 salt bridge). · RESULTS; DISCUSSION

  4. SuggestedPSP-F case 2, diagnosed as PSP, instead had a GPT fold intermediate between GGT and PSP; because its filaments differed from both, the authors propose a distinct disease, limbic-predominant neuronal inclusion body 4R tauopathy (LNT), with spherical limbic 4R neuronal inclusions.

    Two GPT filament types (one protofilament; paired protofilaments enclosing a large solvent-filled cavity); Krios G4 data (cold FEG, Falcon-IV) resolved two GPT type 1 main-chain conformations, one at 1.9 Å with 19 ordered waters per asymmetric unit; GGT-like substructures at residues 356-378 and 273-294 plus 312-346. · RESULTS; DISCUSSION; Clinical history and neuropathology; Figure legends

  5. SuggestedFolds support a hierarchical classification: ordered-core extent tracks isoform composition (3R+4R, 3R, 4R), and 4R folds split into three-layered (PSP, GGT, GPT) and four-layered (CBD, AGD) classes; familial British and Danish dementias (FBD, FDD) share the Alzheimer fold.

    Shared core: R3, R4 and 10-13 C-terminal residues; N-terminal extension is one or two residues for Alzheimer and chronic traumatic encephalopathy folds, over half of R1 for Pick, and all of R2 for 4R folds, which exclude 3R isoforms. FBD and FDD filaments were identical to the paired helical filaments of Alzheimer's disease and primary age-related tauopathy. · DISCUSSION; RESULTS

Methods, data and reproducibility

Sarkosyl-insoluble tau filaments from autopsy brain regions were imaged by cryo-EM and solved by helical reconstruction in RELION, with atomic models built in COOT and refined in REFMAC/PHENIX; folds were compared to prior structures and grouped by ordered-core extent and layering, alongside immunoblots, immunohistochemistry and silver staining.

  • PSP brains: three PSP-RS, two PSP-F (case 2 later designated LNT), one PSP-P, one PSP-CBS; frontal, temporal or entorhinal cortex, thalamus, putamen · human · n = seven individuals
  • GGT frontal cortex: GGT-I and GGT-II structures solved; GGT-III filaments lacked twist and were not solved · human · n = one case each of GGT-I, GGT-II and GGT-III
  • Stage 3 AGD nucleus accumbens; both cases also held Alzheimer-type filaments, case 1 additionally CTE type I filaments · human · n = two individuals
  • ARTAG hippocampus (majority filament species solved), with Alzheimer PHFs and SFs also present · human · n = one individual
  • Inherited 4R tauopathy, MAPT intron 10 +3 (MSTD family history) or +16 (familial frontotemporal dementia) mutation; frontal cortex · human · n = two +3 cases, one +16 case
  • FBD (ITM2B stop-codon loss) hippocampus and FDD (ITM2B decamer duplication) temporal cortex · human · n = one FBD case, one FDD case
  • Previously reported tau folds (Alzheimer's disease, CTE, Pick's disease, CBD, PART) used as structural comparators · human
Methods & tools
Cryo-EM: Thermo Fisher Titan Krios and Krios G4 (Gatan K2/K3, Falcon-IV detectors), Thermo Fisher EPU with aberration-free image shift (AFIS), RELION (motion correction, helical reconstruction, Bayesian polishing, CTF refinement), CTFFIND-4.1, crYOLO, COOT, MOLPROBITY, REFMAC, PHENIX, Sarkosyl-insoluble tau extraction and immunoblotting (anti-tau T46), Immunohistochemistry (AT8, RD3, RD4) and Gallyas-Braak silver staining, Whole-exome sequencing (Agilent SureSelectTX V6; Illumina HiSeq4,000)
Reagents & environment
Sarkosyl extraction (10 mM Tris-HCl pH 7.5, 0.8 M NaCl, 10% sucrose, 1 mM EGTA, 2% sarkosyl); antibodies T46 (1:2,000), RD3, RD4 (1:1,000) and AT8 (1:300); Quantifoil Au R1.2/1.3 grids; RELION, CTFFIND-4.1, crYOLO, COOT, MOLPROBITY, REFMAC, PHENIX.
Experimental conditions
Titan Krios microscopes with Gatan K2 or K3 detectors and GIF-quantum filter (20 eV slit); PSP-F case 2 on a 300 keV Krios G4 (cold FEG, Selectris X filter at 10 eV, Falcon-IV, 50 e^/Å^2, 300 images/h, EER); Vitrobot Mark IV; HiSeq4,000 for exomes.
Reproducibility gaps
EMDB, PDB and EMPIAR accession numbers are unfilled placeholders (EMD-xxxx, XXXX) in this preprint; other data on reasonable request to corresponding authors; no code availability statement; most software versions unstated.
Limitations (6)
  • No structure was obtained for the GGT-III case because its filaments lacked twist, so whether GGT-III is a separate disease from GGT-I and GGT-II remains a suggestion. · stated by the authors
  • AGD type 3 filaments were not resolved well enough for atomic modelling; their asymmetric two-protofilament packing is inferred only from low-resolution cross-sections. · stated by the authors
  • Even the 1.9 Å GPT map could not reveal what the non-proteinaceous densities are, probably because these molecules do not follow helical symmetry and are averaged out; their identity in all folds remains open. · stated by the authors
  • Sample size was not predetermined statistically, and experiments were neither randomized nor blinded. · stated by the authors
  • LNT is defined from a single individual (PSP-F case 2) sampled in one brain region; whether the GPT fold recurs in other patients with similar limbic spherical 4R inclusions, or elsewhere in this brain, was not examined. · noted in this summary
  • Most entities rest on one or two brains (single GGT-I, GGT-II, ARTAG, FBD and FDD cases), largely one region each; the link from relative 4R tau overproduction to the AGD fold is inferred from three mutation carriers without quantifying 4R:3R expression or testing exon 10 mutants. · noted in this summary

Read this if you study tauopathies or tau aggregation and want residue-level cryo-EM folds for PSP, GGT, AGD, ARTAG and MAPT intron 10 cases, or a fold-based scheme for classifying tau diseases.

Citations
879
FWCI
94.14
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISFalcon '18Falcon '18Zhang '19Qi '23Rana '26Lövestam '22
80% semantic matchNeuroscience
Benjamin Falcon, Wenjuan Zhang, Manuel Schweighauser+6 more·Acta Neuropathologica·Oct 1, 2018· 362 citations