. Cryo-EM structures of amyloid-β 42 filaments from human brains. Science. 2022 Jan 14;375(6577):167-172. Epub 2022 Jan 13 PubMed.


Please login to recommend the paper.


  1. This impressive new research from the Cambridge group supports previous NMR findings of Tycko and coworkers (Lu et al., 2013; Qiang et al., 2017) of a predominant Aβ structure in a particular AD brain. Using amyloid-binding dyes such as luminescent conjugated oligothiophenes (LCOs), we also found that the dense, amyloid cores of plaques are quite similar among cortical regions of a given brain, but that the LCO spectra differ among individuals and among familial and sporadic AD cases (Rasmussen et al., 2017). (Note, however, that the present study likely analyzed some of the same familial brains as those in Rasmussen et al., and thus the findings of the two studies are not completely independent.)

    LCOs and cryo-EM do not similarly assess the structural features of polymerized Aβ. LCOs (particularly when used in combinations) largely show the arrangement and packing of Aβ-amyloid fibrils; the readout varies with the different Aβ40/42 ratios, and it also is sensitive to age-related changes in the aggregates (Nyström et al., 2013). LCOs also are not specific for Aβ, in that they bind to a variety of amyloid structures (e.g., Aslund et al., 2009). Using two different LCOs, we found within a given brain polymorphic Aβ-amyloid cores that cluster as clouds around a dominant structure (Rasmussen et al., 2017). Furthermore, the dominant structures of the plaques in human AD cases could be partially recapitulated in experimental mice in our in vivo seeding paradigm.

    An important finding of the present and previous (Kollmer et al., 2019) cryo-EM investigations is that synthetic Aβ fibrils have a different structure compared to brain-derived Aβ fibrils. This observation urges caution in the interpretation of analyses of synthetic Aβ assemblies, and it may explain why synthetic Aβ is a poor seed compared to brain-derived material (Meyer-Luehmann et al., 2006). However, whether this poor seeding is simply based on the molecular structure of the assemblies or on brain-derived cofactors is still an open question. It will be exciting to investigate the comparative seeding potential of Aβ fibrils that have been well-characterized by cryo-EM in animal models. 

    The gorgeous and informative cryo-EM images of Aβ assemblies presented by Yang and colleagues are a significant step forward in deciphering the link between protein aggregation and disease. As always in biology, such nice findings also raise more questions. For example, what role might age alone play in defining the molecular architecture of Aβ fibrils? The present analysis focuses on AD patients with end-stage disease, at least two (maybe three) decades after Aβ deposition has started in the brain; the highest specific Aβ seeding activity, however, has been found at early stages of the pathogenic process (Aoyagi et al., 2019; Ye et al., 2017). 

    Does the sarkosyl extraction procedure favor the isolation of specific Aβ structures? It will be useful to know how mutations that alter the amino acid sequence of Aβ influence the structure of the fibrils (and, for that matter, how mutations outside of Aβ—in APP and presenilin—manage to promote different structures). And how do Aβ40, Aβ42 and other variants interact in fibril formation, and what are the consequences for disease? Does fibril structure influence the sensitivity of imaging agents (or assays) for quantifying Aβ load in the brain? By furnishing new insights into protein aggregation, cryo-EM will be highly useful in addressing these and other open questions in the field.


    . Molecular Structure of β-Amyloid Fibrils in Alzheimer's Disease Brain Tissue. Cell. 2013 Sep 12;154(6):1257-68. PubMed.

    . Structural variation in amyloid-β fibrils from Alzheimer's disease clinical subtypes. Nature. 2017 Jan 12;541(7636):217-221. Epub 2017 Jan 4 PubMed.

    . Amyloid polymorphisms constitute distinct clouds of conformational variants in different etiological subtypes of Alzheimer's disease. Proc Natl Acad Sci U S A. 2017 Dec 5;114(49):13018-13023. Epub 2017 Nov 20 PubMed.

    . Evidence for Age-Dependent in Vivo Conformational Rearrangement within Aβ Amyloid Deposits. ACS Chem Biol. 2013 Mar 29; PubMed.

    . Novel pentameric thiophene derivatives for in vitro and in vivo optical imaging of a plethora of protein aggregates in cerebral amyloidoses. ACS Chem Biol. 2009 Aug 21;4(8):673-84. PubMed.

    . Amyloid polymorphisms constitute distinct clouds of conformational variants in different etiological subtypes of Alzheimer's disease. Proc Natl Acad Sci U S A. 2017 Dec 5;114(49):13018-13023. Epub 2017 Nov 20 PubMed.

    . Cryo-EM structure and polymorphism of Aβ amyloid fibrils purified from Alzheimer's brain tissue. Nat Commun. 2019 Oct 29;10(1):4760. PubMed.

    . Exogenous induction of cerebral beta-amyloidogenesis is governed by agent and host. Science. 2006 Sep 22;313(5794):1781-4. PubMed.

    . 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.

    . Aβ seeding potency peaks in the early stages of cerebral β-amyloidosis. EMBO Rep. 2017 Sep;18(9):1536-1544. Epub 2017 Jul 12 PubMed.

    View all comments by Lary Walker
  2. This is an interesting and important study. The authors report the first structure of Aβ42 fibrils from the brain and although there have been other reports of Aβ42 fibril structures assembled from monomer in vitro, previous work has indicated that the structures of fibrils isolated from brain are quite different.

    Another interesting insight is that the authors observed only two different structures in 10 human brain samples from a range of disease phenotypes, including people with familial AD (FAD) mutations. Type I is found in sporadic AD (sAD), while type II is a minor component in sAD and the dominant form in FAD. Type II was found exclusively in a variety of different disease subtypes, such as PDD, DLB, and FTD. For tau, α-synuclein, and TDP43 amyloids, different structural polymorphs are associated with disease subtypes, and with many disease subtypes in AD, it is possible that there may have been more structures found, which would complicate strategies to image and target the different structures.

    Other human brain amyloid structures have been reported from the Tycko (Ghosh et al., 2021) and Fändrich (Kollmer et al., 2019) laboratories, and Collinge and Tycko (Ghosh et al., 2021; Qiang et al., 2017) have published evidence that fibrils from non-demented individuals and fibrils from the rapidly progressing form of AD display subtle structural differences by solid-state NMR compared to fibrils from sAD. It is possible that these distinct structures in human brain are associated with different brain locations and disease phenotypes than the ones sampled by Scheres, Goedert, and co-workers, but it is also possible that some of the differences are due to methodology. In order to observe single fibrils in cryo-EM, Scheres and Goedert used Sarkosyl detergent extraction. Tycko and Fändrich used milder, less denaturing conditions, but in order to observe single fibrils Tycko used sonicated brain amyloid to seed Aβ40 monomer addition. Fändrich used amyloid isolated from leptomeninges, which is more soluble. It is interesting that several of the structures are very similar in the region between amino acids 25-35, suggesting that this region is the least polymorphic. Ironically this also represents the “toxic domain” that was the poor man’s Aβ substitute and widely used in the 1990s (Behl et al., 1992).   

    Amyloid deposits in human brain show differences in monoclonal antibody and fluorescent dye binding. If there are only two structures in sAD, then differences in dye and antibody binding are either due to polymorphic structures in the amino-terminal “fuzzy coats” (residues 1-12) of the type I and type II structures, and of many in vitro structures that are insufficiently ordered to be discerned in the cryoEM structure. Alternatively, dye and antibody bonding may be influenced by other proteins or molecules that associate with fibrils. This amino terminal domain contains the major binding site for most anti-Aβ antibodies. These antibodies recognize different epitopes and have different binding modes to the same sequence within this region, suggesting that this region can adopt different structures in vivo (Reyes-Ruiz et al., 2020).

    The big questions remaining include: What structures do brain Aβ oligomers adopt? Are they distinct from that of fibril protofilaments? The most obvious possibility is that some oligomers could be small pieces of the fibrils or fibril protofilaments (Wu et al., 2010), but there is immunological evidence that other structural classes of oligomers also exist in brain (Glabe, 2008). 


    . Molecular structure of a prevalent amyloid-β fibril polymorph from Alzheimer's disease brain tissue. Proc Natl Acad Sci U S A. 2021 Jan 26;118(4) PubMed.

    . Cryo-EM structure and polymorphism of Aβ amyloid fibrils purified from Alzheimer's brain tissue. Nat Commun. 2019 Oct 29;10(1):4760. PubMed.

    . Structural differences in amyloid-β fibrils from brains of nondemented elderly individuals and Alzheimer's disease patients. Proc Natl Acad Sci U S A. 2021 Nov 9;118(45) PubMed.

    . Structural variation in amyloid-β fibrils from Alzheimer's disease clinical subtypes. Nature. 2017 Jan 12;541(7636):217-221. Epub 2017 Jan 4 PubMed.

    . Vitamin E protects nerve cells from amyloid beta protein toxicity. Biochem Biophys Res Commun. 1992 Jul 31;186(2):944-50. PubMed.

    . An "epitomic" analysis of the specificity of conformation-dependent, anti-Aß amyloid monoclonal antibodies. J Biol Chem. 2020 Dec 9; PubMed.

    . Fibrillar oligomers nucleate the oligomerization of monomeric amyloid beta but do not seed fibril formation. J Biol Chem. 2010 Feb 26;285(9):6071-9. PubMed.

    . Structural classification of toxic amyloid oligomers. J Biol Chem. 2008 Oct 31;283(44):29639-43. PubMed.

    View all comments by Charles Glabe
  3. This is excellent work, demonstrating the ability of cryo-EM to determine detailed molecular structures of 42-residue Aβ fibrils that were extracted directly from cortical tissue. The close resemblance of peptide conformations in these brain-extracted fibrils to conformations in “synthetic” 42-residue Aβ fibrils, determined six or seven years ago by the Ishii, Meier, and Griffin labs with solid-state NMR methods, is especially interesting to me. Although the conformations are very similar, the details of some side chain orientations and side chain-side chain interactions are different. This shows how subtle the structural distinctions among amyloid polymorphs can be. It is amazing how many similar but different structures these peptides can form.

    One caveat is worth considering: Although the cryo-EM-based structures are certainly correct, it is difficult to know whether these are the only or the predominant polymorphs in the original tissue, since fibrils extracted from cortical tissue tend to be highly self-associated, entangled with one another, and coated with other material. The cryo-EM analysis focuses on fibrils that are relatively “clean,” and a large fraction of the fibril segments end up being discarded. Therefore, the relative abundances in the original tissue cannot be inferred readily from the cryo-EM results. In my lab’s earlier studies of 42-residue Aβ fibrils from cortical tissue of Alzheimer’s disease patients, which were based on solid-state NMR spectra of isotopically labeled fibrils that were seeded with amyloid-containing tissue extracts, we concluded that there were two predominant 42-residue polymorphs, possibly related to the two polymorphs in this paper from Goedert and coworkers (Qiang et al., 2017). However, the solid-state NMR spectra also showed signals from other polymorphs, with variable intensities in samples from different patients.


    . Structural variation in amyloid-β fibrils from Alzheimer's disease clinical subtypes. Nature. 2017 Jan 12;541(7636):217-221. Epub 2017 Jan 4 PubMed.

    View all comments by Robert Tycko
  4. Where’s Wally?

    The avalanche of cryo-EM studies in neurodegenerative diseases continues. This latest offering from the Goedert laboratory highlights the S-shaped protofilament fold in Aβ extracted from human brain tissue. The method of extraction appears important, as they report that a modified sarkosyl extract (adding sarkosyl following homogenization), used previously for isolating α-synuclein, provides “abundant Aβ filaments alongside other amyloids,” whereas a more standard sarkosyl extract (adding sarkosyl at later stages) yields “only tau filaments.” This may suggests that sarkosyl extraction protocols are less reproducible than expected, even though there are reports (e.g, Diner et al., 2017) that show that the “sarkosyl in homogenate” approach gives a very similar result as a traditional multistep approach.

    The bigger issue is how to select an Aβ fibril as opposed to a tau fibril for cryo-EM analysis. Typical neurofibrillary tangle (NFT) paired helical filaments (PHF) are clearly distinguishable from single straight or twisted filaments. These straight filaments measuring 10 nm are more likely to be Aβ amyloid than tau tangles/neurites, at least based on standard electron microscopy of fixed sectioned brain. We know that a sarkosyl preparation contains a large amount of Aβ as well as tau (mostly N-terminally truncated). Looking at a negatively stained preparation of a sarkosyl extract then becomes a bit of a “Where’s Wally?” issue. Which structures are tau, which are Aβ, and what’s in all the non-filamentous aggregates in the background that are rarely mentioned? Be that as it may, the cryo-EM images (Fig S1A) look like entirely ~10 nm thick Aβ42 filaments, but concurrent negative EM immunostains showed (Fig S8) filaments of >20 nm diameter according to the 50 nm scale bar and 10 nm gold particles. Are these tau filaments with fuzzy Aβ coats decorated with anti-Aβ antibodies?

    Notwithstanding these issues, the study describes for the first time the structure of the core (9/12-42) of Aβ42 fibrils extracted directly from 10 postmortem brain samples, including those from subjects who died with sporadic AD, familial AD, or other forms of neurodegenerative disease or aging with AD comorbidity. The authors identified two distinct types of fibril structures, each featuring essentially the same S-shaped core protofibrils, similar to those observed before in in vitro experiments (Fig 4B). However, protofilaments were assembled into fibrils differently. While Type I fibrils predominated in subjects with sporadic AD, type II fibrils predominated in those with familial AD as well as other neurodegenerative diseases. Both fibrillar structures differed from those of fibrils assembled in vitro. Interestingly, APP knock-in mice had Aβ fibrils with a type II structure.

    Since the ordered cores of amyloids are assumed to provide templates for autocatalytic growth, the structures of the ordered cores of amyloids may define “conformers” (Alzforum Comments: What is a conformer? by Scheres and Goedert, 07 Jan 2020). Considering the S-shape as part of the topology of a Greek key motif, we have noticed that this Greek key motif is a key element observed so far in all transmissible amyloids including Aβ42,  α-synuclein (Li et al., 2018; Tuttle et al., 2016), and PrP (Kraus et al., 2021). We also found elements of this “transmissible conformer” in the “oligomeric” prefibrillar structure of the Aβ17-42 tetramer (Streltsov et al., 2011). But it has not been seen in any of the secondary tau cryo-EM structures published to date. This idea is seen in the figure below.


    . Enrichment of Detergent-insoluble Protein Aggregates from Human Postmortem Brain. J Vis Exp. 2017 Oct 24;(128) PubMed.

    . High-resolution structure and strain comparison of infectious mammalian prions. Mol Cell. 2021 Nov 4;81(21):4540-4551.e6. Epub 2021 Aug 25 PubMed.

    . Cryo-EM of full-length α-synuclein reveals fibril polymorphs with a common structural kernel. Nat Commun. 2018 Sep 6;9(1):3609. PubMed.

    . Crystal structure of the amyloid-β p3 fragment provides a model for oligomer formation in Alzheimer's disease. J Neurosci. 2011 Jan 26;31(4):1419-26. PubMed.

    . Solid-state NMR structure of a pathogenic fibril of full-length human α-synuclein. Nat Struct Mol Biol. 2016 May;23(5):409-15. Epub 2016 Mar 28 PubMed.

    View all comments by Victor Streltsov
  5. For me, one of the still largely unanswered questions in the Alzheimer’s field is how the addition of two amino acids at the C-terminus of the Aβ peptide results in such a dramatic difference in the progression of AD.

    Aβ42 predominates in the parenchymal plaques in AD patients, while the more abundant Aβ40 is either cleared from the brain or forms amyloid deposits on or near the cerebral vasculature. The structures of fibrils that have emerged of the Aβ40 and Aβ42 peptides, both in vitro and in vivo, have consistently shown that the additional two amino acids allow the formation of a third β-strand and “S-shaped” structure for Aβ42. So, in this sense, the new brain-derived fibril structures are not so different from the in vitro fibrils.

    One of the striking observations in the current brain-derived Aβ42 structures of Yang et al. is that while the S-shaped structure exists, it is different in packing and electrostatic interactions compared to the in vitro structures determined by cryo-EM and NMR. One common feature of the in vitro structures has been a direct electrostatic interaction between the carboxylate of the C-terminal Ala42 and the positively charged Lys28 side chain. Electrostatic interactions between Lys28 and either Glu22 or Asp23 are often observed in Aβ40 fibrils (both in vitro and in vivo) suggesting that this difference in part drives the different folding patterns between Aβ40 and Aβ42.

    The observation of a possible bound metal ion at Glu22-Asp23 in the brain-derived fibrils described by Yang et al. may in part be relevant for how these fibrils fold and, more generally, for the role of metal ions in influencing fibril morphology. These observations then raise the importance of both the environment in which the Aβ peptides fold and possible co-factors that influence folding.

    View all comments by Steven Smith
  6. This is an interesting study from the laboratories of Michel Goedert and collaborators, where the authors, Yang et al., describe molecular structures of Aβ42 filaments in human brain. The structures of Aβ42 filaments in brain have not been well-defined. Here, Aβ42 filaments taken postmortem from the brains of 10 patients were analyzed at high resolution using cryo–electron microscopy (cryo-EM). Five individuals had Alzheimer’s disease, with three sporadic and two familial cases, and the other five individuals had other forms of neurodegenerative disease. Two structurally related S-shaped protofilament folds were identified and classified as two different types. Type I filaments were found primarily in the brains of individuals with sporadic Alzheimer’s disease, while type II filaments were identified in individuals with familial Alzheimer’s disease and other forms of neurodegenerative disease.

    The study shows that both type I and II structures differed from those of fibrils produced in vitro. This observation strongly suggests, maybe as expected, that the environment has a major impact on the folding pathway from the monomeric state to fibrillar Aβ forms. Even different in vivo conditions (e.g., sporadic versus familial Alzheimer’s disease) seem to have a direct impact on the final fibrillar forms of Aβ in neurodegenerative disease.

    Intriguingly, the AppNL-F knock-in mouse model, which expresses humanized Aβ, was shown to possess Aβ identical to type II filaments isolated from human brains. However, it might be too early to tell whether this preclinical Alzheimer’s disease model is more representative of familial Alzheimer’s disease based on these data alone. In addition to the AppNL-F knock-in model, it would be of great value to characterize the Aβ fibril species in other commonly used Alzheimer’s disease models. Observations like these may have important translational value in developing anti-Aβ agents.

    Yang et al. focus on the sarkosyl-insoluble fraction of Aβ from human brain. As sarkosyl is stringent enough to solubilize most folded proteins in the brain, less robust oligomeric forms of misfolded Aβ might be lost. Thus, different types of soluble Aβ forms, such as oligomers and protofibrils, could be excluded in the preparation and thus not subject to investigations in this cryo-EM study. Different studies suggest that soluble Aβ forms such as oligomers and protofibrils may represent the primary pathological species in Alzheimer’s disease (Nilsberth et al., 2001; Walsh et al., 2002). These species have proven notoriously difficult to investigate in detail with currently available technologies. The dynamics, stability, and transient lifetime of these soluble toxic species hamper the possibility to precisely pinpoint the toxic structural aspects.

    These types of studies could, in the future, add to our understanding of how the different binding profiles of clinical investigated anti-Aβ agents exercise their therapeutic effect. However, it is important to appreciate that these described Aβ structures probably represent a small fraction of all forms of Aβ in the human brain. They most likely belong to the forms of Aβ late in the aggregation pathway.

    These identified structures of Aβ from the human brain could also be used to assess the binding profiles of different Aβ PET tracers, and to develop new PET ligands. For instance, it is speculated that the PET tracer PiB probably labels both type I and II filaments as it visualizes Aβ deposits in both sporadic and familial Alzheimer’s disease. It is worth remembering that the PET ligand PiB does not detect the diffuse plaques in brains of the Arctic Alzheimer’s disease mutation carriers, where glutamic acid is changed for glycine in position 22 of Aβ, leading to a propensity to form Aβ protofibrils and Alzheimer’s disease (Schöll et al., 2012). 

    A bit surprising is that there were no major differences described between the five Alzheimer’s disease cases and the five cases with other neurodegenerative diseases. In the latter five cases, Aβ levels should be much lower in the brain. This does not seem to have an impact on the cryo-EM findings.


    . The 'Arctic' APP mutation (E693G) causes Alzheimer's disease by enhanced Abeta protofibril formation. Nat Neurosci. 2001 Sep;4(9):887-93. PubMed.

    . Naturally secreted oligomers of amyloid beta protein potently inhibit hippocampal long-term potentiation in vivo. Nature. 2002 Apr 4;416(6880):535-9. PubMed.

    . Low PiB PET retention in presence of pathologic CSF biomarkers in Arctic APP mutation carriers. Neurology. 2012 Jul 17;79(3):229-36. PubMed.

    View all comments by Lars Lannfelt

Make a Comment

To make a comment you must login or register.

This paper appears in the following:


  1. Cryo-EM Unveils Distinct Aβ42 Fibril Structures for Sporadic, Familial AD