In recent years, support has grown for the idea that the same aggregating protein can fold up into different toxic shapes, giving rise to distinct neurodegenerative diseases. Now, in the August 29 Nature, researchers led by Michel Goedert and Sjors Scheres at the MRC Laboratory of Molecular Biology in Cambridge, England, in collaboration with Bernardino Ghetti and colleagues at Indiana University School of Medicine in Indianapolis, provide the first molecular-level proof that this happens.

  • Cryo-EM reveals the molecular structure of tau filaments from a person with Pick’s disease.
  • The molecule folds in a completely different way than it does in AD.
  • The results may enable design of more specific reagents.

Using cryo-electron microscopy, they detailed the structure of tau filaments taken from the brain of a person who died of Pick’s disease. Intriguingly, these fibrils contained a fold quite different from that seen in cryo-EM of tau tangles in Alzheimer’s disease. Pick’s disease tau folded up into a J shape, instead of the C shape seen in AD tangles. Domains aligned in a different configuration as well, exposing a distinct set of residues on the surface. The findings explain some known differences in the composition of deposits in Pick’s and AD brains, and may enable the design of more specific reagents for each disease, the authors said. In particular, most current tau PET tracers do not recognize Pick’s deposits. “This provides a framework for designing specific tracers for each filament,” Goedert told Alzforum.

Others greeted the findings with enthusiasm. “By showing that tau filaments in Pick’s disease differ from those in Alzheimer's disease, these intriguing findings provide direct evidence that tau adopts different three-dimensional architectures, or proteopathic strains, in different diseases,” Lary Walker and David Lynn at Emory University, Atlanta, wrote to Alzforum (full comment below).

Different Strokes. In Pick’s disease (left), tau molecules adopt a long J-shaped fold, while in AD (right) they form a C that contains fewer domains. [Courtesy of Falcon et al., Nature.]

The findings have far-reaching implications, noted Byron Caughey at the National Institutes of Health in Hamilton, Montana. “The fact that tau filaments of Pick’s and Alzheimer’s diseases have distinct amyloid core folds implies that they will likely have different chemical surfaces, tendencies to interact with other factors and tissue components, preferred sites of accumulation, cytotoxicities, and thereby, neuropathological and clinical consequences,” he wrote to Alzforum (full comment below).

Scheres and Goedert previously solved the three-dimensional structure of Alzheimer’s tau. Tau molecules come in different isoforms, some of which contain three-repeat domains (3R tau), and some four-repeat (4R tau). AD filaments were known to equally incorporate 3R and 4R varieties. Cryo-EM revealed the reason, finding that the C-shaped core of the aggregate comprised the third and fourth repeat, which are present in all isoforms—it’s the second repeat that gets spliced out in the 3R isoform. Meanwhile, the free ends of each tau molecule waved around randomly in solution, forming a fuzzy coat around the fibril. Individual monomers stacked up to form filaments. These filaments pair up via hydrogen bonding along their backs (Jul 2017 news). 

In contrast to AD, deposits in Pick’s disease are known to incorporate only 3R tau. To see why this might be, first author Benjamin Falcon isolated aggregates postmortem from the frontotemporal cortex of a woman with Pick’s disease. Like AD fibrils, the filaments were surrounded by a disordered, fuzzy coat made up of the free ends of the tau molecules. The researchers removed this with pronase treatment, and then used cryo-EM to map the structure of the aggregated protein core to 3.2 ångstrom resolution.

They found that the core consisted of residues 254 to 378 of 3R tau, making it longer than the AD core. This section includes repeats 1, 3, and 4. The core contained nine distinct β-strands; two from repeat 1, three from R3, three from R4, and one just past R4. The molecule kinks at Cys322 in R3 just after the fourth β-strand, creating a hairpin turn that enables the Pick’s fold (see image above). This allows β4/β5, β3/β6, β2/β7, and β1/β8 to pair up alongside each other, with β9 folding back onto the backside of β-8 along the short arm of the J.

This structure solves the riddle of why Pick’s deposits exclude 4R tau, the authors noted. For 4R tau to assume this J configuration, the R2 domain would have to replace R1. However, the R2 domain contains several side chains that would not pack properly into this structure. For example, glutamine at position 269 of 3R tau would be replaced by a valine, but the latter residue sports a branched side chain that would not fit. The authors confirmed the isoform selectivity in vitro, finding that the filaments used in this study seeded aggregation of 3R tau, but not 4R.

The Pick’s fold also explains why tau deposits in this disorder are not phosphorylated at Ser262. This residue packs into the interior of the Pick’s core at the first tight turn. In AD tau, Ser262 is exposed. The researchers saw other notable differences as well. While Cys322 is exposed in the Pick’s fold, it snuggles into the interior of the molecule in the AD fold. Conversely, Asp348, located in the middle of an R4 β-strand, pokes into solution in the AD fold, allowing the molecule to bend there and form one end of the C. Asp348 stays interior in the Pick’s fold, and the β-strand travels straight.

As in AD, the folded tau molecules stack up to form filaments, and these come in two varieties. Narrow Pick’s filaments consist of a single strand of stacked tau molecules, while in wide filaments, two strands bind each other at the hairpin turn through weak electrostatic interactions. i.e., van der Waals forces. This differs from AD, where tau filaments are always paired, but can form two filament types depending on their exact configuration.

Do all cases of sporadic Pick’s disease harbor this same structure? Goedert acknowledged that this is not yet known, but preliminary evidence suggests they might. The researchers analyzed tau aggregates from eight more Pick’s patients by immunogold EM. This technique uses antibody binding to characterize molecules, but cannot see fine details of structure. Deposits from all eight brains resembled those from the brain used for cryo-EM in that antibodies to tau’s R1, R3, and R4 domains did not recognize the aggregates, supporting the idea that all these domains were packed into the fibril core and inaccessible for antibody binding. When deposits were denatured and run on a gel, all three antibodies did recognize Pick’s tau. The researchers have not yet examined any Pick’s cases with mutations in tau. It is possible these could form distinct structures, Goedert noted.

The Pick’s and AD structures shed no light on what makes tau fold into one shape or the other. In solution, tau usually remains soluble, but can be triggered to aggregate by adding the polysaccharide heparin. Perhaps there are co-factors in neurons that induce tau to fold in particular ways, Scheres suggested. Alternatively, post-translational modifications such as phosphorylation may guide folding. For their part, Walker and Lynn wondered if smaller assemblies of tau push aggregation in one direction versus another. “Is the molecular architecture of fibrils related in some way to the formation, structure and/or pathogenicity of oligomers?” they asked.

Some of these possibilities could be investigated in mice that express human tau. Researchers could alter individual amino acids and see the effect on folding, providing clues to the earliest events in disease. First, however, Goedert and Scheres are continuing to investigate tau deposits in people. They are using cryo-EM to analyze co-aggregates of tau and PET tracers such as AV1451 to find out where these ligands bind. They are also attempting to isolate enough tau filaments from people with progressive supranuclear palsy and corticobasal degeneration to do cryo-EM. These deposits contain only 4R tau, and might assume a third configuration, distinct from the Pick’s and Alzheimer’s tau.—Madolyn Bowman Rogers


  1. Cryo-electron microscopy (cryo-EM, aka electron cryomicroscopy) has become the go-to tool for determining the molecular architecture of alternatively folded proteins in various disease states. A particular advantage of cryo-EM is that it can be applied to individual protein polymers taken directly from the brain, and thus yield a reconstruction of the protein's molecular shape as it existed in real life. Falcon and colleagues previously exploited this capability to ascertain the structure of the ordered core of tau protein in Alzheimer's disease (Fitzpatrick et al., 2017). Now they show that polymerized tau in Pick's disease, which (unlike in AD) consists only of the three-repeat isoform of tau, adopts a disease-specific fold (called the “Pick fold”). The brain-derived, Pick-folded tau is able to seed the self-assembly of recombinant 3R- but not 4R-tau, explaining why only 3R-tau is present in Pick bodies even though humans express approximately equal amounts of both isoforms.

    By showing that tau filaments in Pick’s disease differ from those in Alzheimer's disease, these intriguing findings provide direct evidence that tau adopts different three-dimensional architectures, or proteopathic strains, in different diseases. Function follows form, then, in pathobiology as in normobiology, but the next step is to determine exactly how molecular architecture is linked to the different types of disease caused by a given protein. It will be important to fit oligomeric assemblies into the puzzle—for example, is the molecular architecture of fibrils related in some way to the formation, structure, and/or pathogenicity of oligomers? And finally, how can detailed structural information guide therapeutic strategies? This cryo-EM analysis by Falcon and co-workers, and the many that certainly will follow, should help to resolve important questions about how aberrant proteins cause disease.

    —David Lynn at Emory is the co-author of this comment.


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

  2. The recent structures from Goedert and colleagues are quite illuminating and have far-reaching implications. The fact that tau filaments of Pick’s and Alzheimer’s diseases have distinct amyloid core folds implies that they will likely have different chemical surfaces, tendencies to interact with other factors and tissue components, preferred sites of accumulation, cytotoxicities, and thereby, neuropathological and clinical consequences. The results also provide a structural basis for the profound amyloid seeding/templating selectivity that we have observed for Pick’s disease tau aggregates in our ultrasensitive 3R tau RT-QuIC assay (Saijo et al., 2017). The distinct seeding templates presented by Pick’s and AD tau filaments presumably underpins the faithful prion strain-like propagation of these pathological conformers of tau amyloids.


    . Ultrasensitive and selective detection of 3-repeat tau seeding activity in Pick disease brain and cerebrospinal fluid. Acta Neuropathol. 2017 May;133(5):751-765. Epub 2017 Mar 14 PubMed.

  3. Benjamin Falcon’s recent paper describing the crystal structure of three-repeat tau isolated from the brains of Pick’s disease patients is remarkable for the fact that we can now see how the two major isoforms of tau protein fold in distinctly different ways. T-tau (tubulin-associated unit) is a major neuronal cytoskeletal protein found in the CNS encoded by the gene MAPT. Alternative splicing can generate six different isoforms that can be distinguished based on the presence of zero, 1, or 2 N-terminal inserts and the presence or absence of repeat domain 2 (3R or 4R). These three-repeat (3R) or four-repeat (4R) tau species are differentially expressed in neurodegenerative diseases with progressive supranuclear palsy and Alzheimer’s disease primarily expressing the 4R tau isoform while Pick’s disease and behavior variant frontotemporal dementia (bvFTD) primarily express the 3R tau isoform (Ginsberg et al., 2006). 

    This recent finding shows by cryo-electron microscopy the difference in folding between the 3R tau and the 3R/4R tau. Although the 3R and 4R tau proteins are found in equal molar concentrations in healthy neurons, the 4R tau preferentially accumulates in Alzheimer’s disease, and the 3R tau preferentially accumulates in Pick’s disease and some forms of FTD. To date, most approaches to tau immunotherapy have focused on total tau or the 4R tau protein. This study shows us how the two isoforms significantly differ in their folding in the pathogenic state. It will be interesting to see if the 3R tau and 4R tau proteins in the healthy brain retain the same folding structures as observed in this study.

    Among the many functions attributed to tau protein in axons, directing axonal transport has been the subject of recent interest with respect to tau pathology. Anterograde and retrograde transport of cargo in the axon are mediated by the kinesin and dynein motors interacting with microtubules. Tau regulates axon transport directionality by interacting with microtubules and competing for binding of these cargo transport motors. Recently, Lacovich et al. reported that the ratio of 3R tau and 4R tau in the axon can direct the transport of cargo, with higher levels of 3R tau favoring anterograde transport toward the cell body and inhibiting the retrograde transport of those same cargos (Lacovich et al., 2017). Thus, the differences in folding of the C-terminus of the protein of 3R tau and 4R tau may play a role in the directionality of cargo transport.

    Another interesting finding from this study was differences in exposure of the traditional serine 262 phosphorylation site often used as a marker of tau accumulation. Falcon et al. show with the cryo-EM structure of 3R tau that that Serine262 is hidden inside the β-sheet core whereas with 4R tau this residue is assessable for phosphorylation. This may explain the differential phosphorylation of these two tau species.


    . Shift in the ratio of three-repeat tau and four-repeat tau mRNAs in individual cholinergic basal forebrain neurons in mild cognitive impairment and Alzheimer's disease. J Neurochem. 2006 Mar;96(5):1401-8. PubMed.

    . Tau Isoforms Imbalance Impairs the Axonal Transport of the Amyloid Precursor Protein in Human Neurons. J Neurosci. 2017 Jan 4;37(1):58-69. PubMed.

  4. The recent structural studies illuminate that Pick’s disease is picky in selecting a repeat to fold into tau aggregates. 

    Falcon and colleagues have demonstrated an exciting difference at the molecular level between Alzheimer’s disease and Pick’s disease (PiD). The finding is an excellent use of cryo-EM, but it was not only cryo-EM. High-quality “classic” neuropathology dissection was necessary, and this was followed up by biochemical filament extraction and crystallization before cryo-EM could generate a model for the (now known to be) PiD-unique “J” structure of tau aggregates. Western blotting determined the absence of the anti-R2 antibody target and confirmed that the structure was exclusively made up of “a novel fold of 3R tau” (Falcon et al., 2018). This is in striking contrast to the aggregates of AD, which include both 3R and 4R tau. However, the striking element of the report is that tau filaments from the Pick’s disease patient’s brain would only seed aggregation of full-length 3R tau into the Pick “J” form, without having any effect on 4R, while 3R and 4R AD aggregates could seed both 3R and 4R tau to fold into the AD “C” form. In other words, a distinct molecular pathology feature of Pick’s is propagated through selective structural seeding, without gene transcription, transcription processing, or translation mechanisms intervening.

    Modern structural and classical protein biochemistry and neuropathology techniques, among others, have converged elegantly in this work. Its value indeed spreads far beyond structural characterization.

    As has been noted, function follows form, and the differences in form may help untangle differences in (dys)function between AD and PiD. Furthermore, by defining a distinct PiD tau subunit structure, Falcon’s group has defined several concrete questions: How does tau end up with that particular structure versus the AD-typical structure? Is the difference determined early in PiD etiology or is there a “general predegenerative condition” that later diverges toward AD versus PiD (versus FTLD, etc.)? How do the PiD versus AD tau aggregate subunits specifically differ in their dysfunctions? Finally, what processes drive a brain to form Pick folds vs. AD folds?

    In short, this work’s influence will touch far beyond reporting a structural novelty of tau in PiD versus AD.


    . Structures of filaments from Pick's disease reveal a novel tau protein fold. Nature. 2018 Sep;561(7721):137-140. Epub 2018 Aug 29 PubMed.

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

Further Reading

Primary Papers

  1. . Structures of filaments from Pick's disease reveal a novel tau protein fold. Nature. 2018 Sep;561(7721):137-140. Epub 2018 Aug 29 PubMed.