Article https://doi.org/10.1038/s41467-026-69377-1 Structural defects in amyloid-β fibrils drive secondary nucleation Jing Hu 1,2, Tom Scheidt 3,4, Dev Thacker 2,5, Emil Axell 2, Elin Stemme2, Urszula Łapińska 6, Stefan Wennmalm7, Georg Meisl 3,8, Samo Curk9, Maria Andreasen 10, Michele Vendruscolo 3, Paolo Arosio 11, Anđela Šarić 9, JeremyD. Schmit12, Tuomas P. J. Knowles 3,13, Emma Sparr 1, Sara Linse 2, Thomas C. T. Michaels 14,15 & Alexander J. Dear 2,14,15 Formation of new amyloid fibrils and oligomers from monomeric protein on the surfaces of existing fibrils is an important driver of many disorders such as Alzheimer’s and Parkinson’s diseases. The structural basis of this secondary nucleation process, however, is poorly understood. Here, we ask whether secondary nucleation sites are found predominantly at rare growth defects: irregularities in the fibril core structure incorporated during their original assembly. We first demonstrate using the specific inhibitor of secondary nucleation, Brichos, that secondary nucleation sites on Alzheimer’s disease- associated fibrils composed of Aβ40 and Aβ42 peptides are rare compared to the number of protein molecules they contain. We then grow Aβ40 fibrils under conditions designed to eliminate most growth defects while leaving the regular fibril morphology unchanged, and confirm the latter using cryo- electron microscopy. We measure both the ability of these annealed fibrils to promote secondary nucleation and the stoichiometry of their secondary nucleation sites, finding that both are greatly reduced as predicted. Re-analysis of published data for other proteins suggests that fibril growth defects may also drive secondary nucleation generally across most amyloids. These find- ings could unlock structure-based drug design of therapeutics that aim to halt amyloid disorders by inhibiting secondary nucleation sites. Amyloid fibrils are characterised by the stacking of proteins into β- sheet-rich aggregates with a hydrophobic core. They can serve func- tional purposes but are also frequently associated with pathology1–4, making the mechanisms that control their proliferation a central pro- blem in disease. Their spontaneous formation requires a slow primary nucleation process producing very short new fibrils from monomeric protein, and their subsequent rapid elongation by monomer addition (Fig. 1a)5. Notably, in many systems, an additional reaction step, sec- ondary nucleation of new fibrils on the surfaces of existing fibrils6–9, dominates new oligomer and fibril production once aggregates appear10,11. Secondary nucleation has been implicated across systems including type II diabetes-associated islet amyloid polypeptide (IAPP)12,13, Parkinson’s disease-associated α-synuclein14, and Alzhei- mer’s disease (AD)-associated tau and amyloid-β (Aβ)10,15,16. Rapid sec- ondary nucleation is characteristic of many disease-associated amyloids, dueboth to its auto-catalytic effect onfibril proliferation and to its ability to produce large quantities of toxic protein oligomers9,17. Analogous surface-catalysed nucleation processes are well known in crystallisation, underscoring the generality of the underlying physics18–20. However, despite its importance, exactly how fibril surfaces cat- alyse secondary nucleation is still not known. A common assumption is that catalysis is uniform across the entire fibril surface, or at least that catalytic sites occur at every plane in the fibril (see Fig. 1b(i)). Received: 29 June 2024 Accepted: 30 January 2026 Check for updates A full list of affiliations appears at the end of the paper. e-mail: alexander.dear@bc.biol.ethz.ch Nature Communications | (2026) 17:1933 1 12 34 56 78 9 0 () :,; 12 34 56 78 9 0 () :,; http://orcid.org/0009-0000-3605-2185 http://orcid.org/0009-0000-3605-2185 http://orcid.org/0009-0000-3605-2185 http://orcid.org/0009-0000-3605-2185 http://orcid.org/0009-0000-3605-2185 http://orcid.org/0000-0002-0185-7730 http://orcid.org/0000-0002-0185-7730 http://orcid.org/0000-0002-0185-7730 http://orcid.org/0000-0002-0185-7730 http://orcid.org/0000-0002-0185-7730 http://orcid.org/0000-0001-6171-9703 http://orcid.org/0000-0001-6171-9703 http://orcid.org/0000-0001-6171-9703 http://orcid.org/0000-0001-6171-9703 http://orcid.org/0000-0001-6171-9703 http://orcid.org/0000-0002-5687-9566 http://orcid.org/0000-0002-5687-9566 http://orcid.org/0000-0002-5687-9566 http://orcid.org/0000-0002-5687-9566 http://orcid.org/0000-0002-5687-9566 http://orcid.org/0000-0003-3593-9248 http://orcid.org/0000-0003-3593-9248 http://orcid.org/0000-0003-3593-9248 http://orcid.org/0000-0003-3593-9248 http://orcid.org/0000-0003-3593-9248 http://orcid.org/0000-0002-6562-7715 http://orcid.org/0000-0002-6562-7715 http://orcid.org/0000-0002-6562-7715 http://orcid.org/0000-0002-6562-7715 http://orcid.org/0000-0002-6562-7715 http://orcid.org/0000-0002-6096-2995 http://orcid.org/0000-0002-6096-2995 http://orcid.org/0000-0002-6096-2995 http://orcid.org/0000-0002-6096-2995 http://orcid.org/0000-0002-6096-2995 http://orcid.org/0000-0002-3616-1610 http://orcid.org/0000-0002-3616-1610 http://orcid.org/0000-0002-3616-1610 http://orcid.org/0000-0002-3616-1610 http://orcid.org/0000-0002-3616-1610 http://orcid.org/0000-0002-2740-1205 http://orcid.org/0000-0002-2740-1205 http://orcid.org/0000-0002-2740-1205 http://orcid.org/0000-0002-2740-1205 http://orcid.org/0000-0002-2740-1205 http://orcid.org/0000-0002-7854-2139 http://orcid.org/0000-0002-7854-2139 http://orcid.org/0000-0002-7854-2139 http://orcid.org/0000-0002-7854-2139 http://orcid.org/0000-0002-7854-2139 http://orcid.org/0000-0002-7879-0140 http://orcid.org/0000-0002-7879-0140 http://orcid.org/0000-0002-7879-0140 http://orcid.org/0000-0002-7879-0140 http://orcid.org/0000-0002-7879-0140 http://orcid.org/0000-0001-8343-9657 http://orcid.org/0000-0001-8343-9657 http://orcid.org/0000-0001-8343-9657 http://orcid.org/0000-0001-8343-9657 http://orcid.org/0000-0001-8343-9657 http://orcid.org/0000-0001-9629-7109 http://orcid.org/0000-0001-9629-7109 http://orcid.org/0000-0001-9629-7109 http://orcid.org/0000-0001-9629-7109 http://orcid.org/0000-0001-9629-7109 http://orcid.org/0000-0001-6931-5041 http://orcid.org/0000-0001-6931-5041 http://orcid.org/0000-0001-6931-5041 http://orcid.org/0000-0001-6931-5041 http://orcid.org/0000-0001-6931-5041 http://orcid.org/0000-0003-3055-607X http://orcid.org/0000-0003-3055-607X http://orcid.org/0000-0003-3055-607X http://orcid.org/0000-0003-3055-607X http://orcid.org/0000-0003-3055-607X http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-69377-1&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-69377-1&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-69377-1&domain=pdf http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-026-69377-1&domain=pdf mailto:alexander.dear@bc.biol.ethz.ch www.nature.com/naturecommunications Nonetheless, several recent studies imply secondary nucleation sites on Aβ fibrils are rare and substoichiometric21–24, casting doubt on this hypothesis, although the stoichiometry of these sites has never been quantified directly. One proposed explanation for this is that catalysis is localised at rare super-spreader fibrils (fibrils with dense secondary nucleation sites), with the remainingmajority of fibrils being incapable of secondary nucleation (see Fig. 1b(ii)). However, recent studies indicate that secondary nucleation sites are spatially isolated fromone another25, that most Aβ42 fibrils have comparable catalytic activity26, and that almost all Aβ42 fibrils possess non-periodic secondary nucleation sites27, rendering this model unlikely for Aβ fibrils. In light of the above findings, it has been proposed that these sites instead correspond to growth defects: defects generated in the fibril core during its initial assembly25. Those types of defect structures that permit the next fibril plane to be assembled with the correct structure on top of the defective fibril plane without imposing an insurmoun- table free energy penalty should exist throughout the fibril length coexisting with otherwise uniform fibril morphology. One example of such a defect is when themonomers in a newfibril plane bind correctly with respect to one another and with the correct fold but laterally offset and potentially also out-of-register from the monomers in the preceding fibril plane during fibril elongation28,29. The next fibril plane can then be assembled in perfect alignment with the offset plane, allowing correct growth to resume (see Fig. 1b(iii)). Small dislocations only partially exposing a filament’s cross-section are possible, as are larger dislocations that lead to entire filaments being misaligned, or even to fibril branching. Other plausible defect structures include (but are not limited to) fibril planes in which one ormore of the monomers is only partially correctly folded, exposing part of the fibril’s hydro- phobic core. Near thermodynamic equilibrium, the stoichiometry of growth defects (the number of defects per monomer in the fibril) should fol- low the Boltzmann distribution, becoming the exponential of the free energy penalty for forming a given type of defect30. Because fibrils are highly stable these penalties are expected to be large, leading to low defect incorporation into growing fibrils. Typically, however, experi- ments use high initial supersaturation (i.e. monomer concentration) to obtain sufficientfibril yield and reaction speed, driving fibril growth far from equilibrium. The average time between successive monomer binding events becomesmuch smaller than the average time taken for an incorrectly bound monomer to detach29. Therefore, many fibril planes arebound sequentially on topof a defectivefibril planebefore it can detach or re-fold. The resultant very strong kinetic trapping can greatly increase defect stoichiometry29,31. Increasing defect incor- poration with increasing supersaturation is a well-documented feature of crystallisation32, of which amyloid formation is often considered a one-dimensional analogue33. Conversely, the common manufacturing technique of annealing can reduce defects in crystalline or pseudo- crystalline materials by using thermal treatment to remove kinetic trapping34–36, suggesting an experimental lever to modulate growth defects. Here, we quantify secondary nucleation site stoichiometries on Aβ42 and Aβ40 fibrils using BRICHOS, a ~100-residue mammalian chaperone that selectively binds to and blocks these sites21,37,38, enabling a direct readout of site numbers per fibril mass. Inspired by annealing, we then use temperature control to growAβ40 fibrils under conditions designed to minimise growth defects whilst preserving their morphology as verified by high-resolution cryo-electron micro- scopy. Brichos binding measurements reveal that their secondary nucleation site stoichiometry is also greatly reduced, and kinetic assays demonstrate that the secondary nucleation rate of such annealed fibrils is reduced commensurately, causally linking defects to secondary nucleation sites. Together, these results identify growth defects as the predominant driver of secondary nucleation on Aβ fibrils, quantify their abundance, and suggest a general, defect-based mechanism for amyloid secondary nucleation supported by thermo- dynamic arguments and re-analysis of prior studies. Results Brichos binds Aβ42 amyloid fibrils tightly with low stoichiometry We first characterised the interaction between the Brichos chaperone domain and Aβ42 fibrils using a microfluidic diffusional sizing (MDS) platform that was previously developed to quantify the interactions between biomolecules in solution under native conditions39,40. In brief, this technique utilises the diffusion profiles of molecules in solution collected at multiple positions as they flow through a microfluidics channel (Fig. 2a). From the analysis of thesediffusionprofiles using the advection-diffusion equation, it is possible to extract the distribution of diffusion coefficients. Brichos molecules that are bound to amyloid fibrils diffuse through the channel cross-section much more slowly than unbound Brichos molecules, exhibiting two distinct diffusion profiles that can be detected and separated mathematically40. Note, this separation becomes inaccurate when one species makes up <10% of the mixture. Alexa-488 labelled proSP-C Brichos was titrated against 24 μM Aβ42 amyloid fibrils at 21 °C. Two different incubation procedures were followed: addition of Brichos to Aβ42 fibrils generated in the absence of chaperone; and incubation of Brichos with monomeric Aβ42 at 37 °C until essentially all monomers had converted into fibrils. In both procedures themixtures were subsequently incubated for 48 h at 21 °Cprior to evaluation of binding. The resulting binding curves are shown in Fig. 2b, c; the data obtained through the two procedures are closely similar. We then fit to these data the Langmuir adsorption isotherm41½Cbound � ½Stot � = ½Cf ree � KD + ½Cf ree � , where KD is the dissociation constant, [Cbound] and [Cfree] are the concentrations of the bound and free cha- perone, respectively. [Stot] is the total concentration of binding sites available on an Aβ42 fibril surface, representing the maximum con- centration of bound molecular chaperone. This procedure yielded KD = 314 nM at 21 °C and [Stot] = 168 nM. Dividing [Stot] by the total concentration of monomer equivalents of Aβ42 in the fibrils revealed the binding stoichiometry s to be about one Brichos molecule per (24μM)/(168 nM)=143Aβ42monomers (Fig. 2c), i.e. s= 1/143. Brichos is predominantly trimeric in solution38, but dissociates into monomers when binding to Aβ fibril secondary nucleation sites22,42. This should not greatly affect the fitted stoichiometry, since each monomer is individually labelled and MDS measures total fluorescence. However, mild self-quenching of trimeric free Brichos is possible, in which case our fitted stoichiometry value would be a mild overestimate. Modulation of Aβ42 aggregation by Brichos reveals rare sec- ondary nucleation sites We next sought to relate our binding measurements to the ability of Brichos to inhibit secondary nucleation of Aβ42 fibrils. To this end, we incubated 3μM of Aβ42 at 21 °C and monitored the time course of amyloid fibril formation by Thioflavin T (ThT) fluorescence in the absence and presence of increasing concentrations of Brichos (Fig. 2d). We observe that the addition of Brichos delays aggregation by an amount that depends on the concentration of the chaperone, in line with previous reports21. To disentangle the effect of Brichos on the rate constants for the individual microscopic steps of aggregation (primary nucleation kn, secondary nucleation k2 and elongation k+), we compared the experimental kinetic profiles with a theoretical chemical kinetics model of aggregation in the presence of Brichos (seeMethods “Kinetic modelling”). In this model, chaperone molecules can bind surface sites of fibrils and in this manner reduce the rate of secondary nucleation, while leaving the rate of the other aggregation steps unaffected43. A prediction from this model based on the unperturbed aggregation rate parameters obtained in the absence of Brichos (left- most curve in Fig. 2d) and the rate parameters determined from the Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 2 www.nature.com/naturecommunications binding experiment shows good agreement with the experimental data. A comparison between the effective rate constant for secondary nucleation k2 with increasing Brichos concentration and the binding curve of Brichos to fibrils confirms that the extent of inhibition of Aβ42 by Brichos closely correlates with the amount of chaperone bound to fibrils (Fig. 2e). Strikingly, we see that almost complete suppression of secondary nucleation (reduction of k2 by >90%, see 2 and 3μMBrichos datapoints) occurs with a Brichos coverage of <1% of the fibrils. We Fig. 1 | Secondary nucleation is a core process driving amyloid fibril pro- liferation; however, its structural basis is not understood. a Overall reaction network governing amyloid fibril formation. Secondary nucleation is often responsible for most new fibril formation. b Possible ways amyloid fibrils might promote secondary nucleation of new fibrils (and oligomers). (i): Uniform catalysis of nucleation across the entire fibril surface. (ii): Rare super-spreader fibrils have dense catalytic sites, with most fibrils being incapable of promoting secondary nucleation. (iii): Secondary nucleation sites are defects in the fibril core, created during fibril elongation. Such defects can coincide with dislocations or branches (LHS, and bottom row), although non-offset defect structures are also possible (RHS). Dark and light blue colours distinguish separate protofilaments. Fibril cross- sections denoted by purple units andmarked by arrows are exposed at dislocation or branching defect sites. The fraction of monomer units exposed may differ from that illustrated here. For simplicity, defects in subsequent figures will be repre- sented schematically as dislocations only. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 3 www.nature.com/naturecommunications thereforeobtain sub-stoichiometric inhibitionof secondarynucleation by Brichos, with almost complete suppression of secondarynucleation obtainedwith only one fibril-bound chaperonemoleculeper about 150 monomers. These results imply that Brichos binding sites and sec- ondary nucleation sites are one and the same. Under these fibril for- mation conditions, secondary nucleation sites therefore have a stoichiometry of s = 1/143, i.e. a frequency of approximately 1 site per 150 monomers in the fibril. With four monomers per plane with two monomers per filament plane and two such filaments in an Aβ42 fibril44,45, this stoichiometry corresponds to one catalytic site per 36 fibril planes on average. A recent evaluation of the density of fibril protrusion during an ongoing Aβ42 aggregation process at pH 6.8 found a distribution peaking at around 1 protrusion per 28 planes, or per 112 monomers, i.e. similar to the rarity of secondary nucleation sites found here at pH 8.046. Fibrils possessing fewer growth defects can be generated with- out measurably altering morphology The low stoichiometry of Aβ42 secondary nucleation sites we have found in this study supports the hypothesis that these sites are rare growth defects in the fibril core structure. To verify this hypothesis, we sought to produce Aβ fibrils under conditions designed to reduce defect stoichiometry by removing their kinetic trapping, so we could subsequently test whether their secondary nucleation sites are also reduced. Removal of supersaturation alone, e.g., by allowing an aggre- gation reaction to proceed to completion, is not expected to remove kinetic trapping of defects. This is because the prevalence of long fibrils (often containing >105 monomers) leads to very slow exchange of protein in the interiors of fibrils with the residual monomeric protein in solution28–30,47,48. The need to depolymerise tens of thousands of monomers from the fibril to access the average defect imposes an insurmountable kinetic free-energy barrier for their removal. Therefore, fibril growth defects will persist in mature fibrils for very long periods of time unless experimental conditions are chosen to almost completely disassemble the fibrils. Also, in situ self-repair of defects is expected to be very slow even upon thermal treatment, since re-folding of defective monomers and alteration of their highly directional bonding to their neighbours is likely impractical without fibril disassembly. Instead, a more practical approach is to prevent kinetic trapping in the first place by growing the fibrils at very low supersaturation (Fig. 3a, b). To do so, we took advantage of the fact that fibril solubility—the monomer concentration at which elongation and dissociation rates equalise—can be altered by changes in temperature. A previous study49 reported a steady increase in Aβ40 solubility with tempera- ture from 37 to 64 °C in 20 mM sodium phosphate, pH 7.4 (Fig. 3c, Supplementary Table 2). We validated this trend by measuring the solubility of Aβ40 at 27, 45, 50, and 60 °C under the same buffer conditions using NMR. In combination with the published solubility value at 37 °C50, these measurements confirmed that Aβ40 solubility increases steadily with temperature from 37 to 60 °C (Supplemen- tary Table 1). An annealing-type procedure involving incubating fibril seeds with a monomeric solution at a concentration slightly above the 60 °C solubility, and gradually reducing the temperature from 60 to 37 °C should then enable fibril growth at almost zero super- saturation (Fig. 3c). Such annealed fibrils are expected to have far fewer growth defects than fibrils produced from a similar (and therefore highly supersaturated) initial monomer concentration at Fig. 2 | Stoichiometry of Brichos binding toAβ42 fibrils is low, and is predictive of secondary nucleation inhibition. a Illustration ofMDS experiment. Mixtures of Alexa-488-labelled proSP-C Brichos chaperone and 24 μM Aβ42 fibrils are intro- duced to the centre of a microfluidic channel. Free Brichos diffuses rapidly toward the channel edges, while fibril-bound Brichos displays slower lateral diffusion and remains therefore localised in the centre of the channel. Top schematic reprinted ("Adapted” or “in part'') with permission from ref. 40. Copyright 2016 American Chemical Society. b The proportion of Brichos bound to fibrils, inferred by math- ematical analysis of the diffusion profiles, decreases with increasing total con- centration C of proSP-C Brichos. By C = 1.5μM it is already below the reliable detection limit of 10%; measurements at higher C would therefore be unin- formative. Sample sizes consisted of 2–3 repeats per condition. c Converting these proportions to free (x axis) and bound (y axis) Brichos concentrations enables global regression of a Langmuir adsorption isotherm to both datasets. This yields KD = 314 ± 85 nM and stoichiometry of 1 binding site per ca. 143 ± 16 Aβ42 mono- mers (R2 = 0.96). Sample sizes consisted of 2–3 repeats per condition. d Kinetic aggregation profiles of 3μMAβ42with 1% seed and increasing total concentrations C of Brichos. Lines indicate global fits using a rate law for Aβ42 with an effective secondary nucleation rate constant k2(C) (see Methods “Kinetic modelling”; adjusted MSE=0.00052). The sample size consisted of 1 repeat per condition. e Decrease in k2(C) with increasing C (red data, left axis) and fractional fibril cov- erage by Brichos (blue data, right axis). Superimposed aremodel projections using parameters determined in c and d selective binding to secondary nucleation sites, k2(C)/k2(0) = 1/(1 + C/KD) (red line; R2 = 0.99), and Langmuir isotherm (blue line; R2 = 0.97). Both closely match the data. Conditions: 20mM sodium phosphate, 0.2mM EDTA, pH 8.0 and 21 ∘C. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 4 www.nature.com/naturecommunications 37 °C. Here, we worked with Aβ40 rather than Aβ42 due to its higher solubility, making it easier to generate sufficient annealed fibril concentrations at low supersaturation. First-generation Aβ40 fibrils were prepared by full aggregation of 57μM Aβ40 monomer at 37 °C. Second-generation annealed and control fibrils were produced by incubating 0.2μM monomer- equivalent of the first generation fibrils with 10μM fresh Aβ40 monomer. To make annealed fibrils, the incubation temperature was started at 60 °C and was reduced by 1 °C every hour until reaching 37 °C (Fig. 3c and Supplementary Fig. 4c). To make control fibrils, the incubation temperature was maintained at 37 °C throughout. Seeding bypasses primary nucleation43,51, and at the low supersaturation used to make annealed fibrils, elongation dominates over secondary nucleation. Therefore, most annealed fibrils are expected to be elon- gated first-generation seeds. Since elongation typically preserves fibril morphology52, using first-generation seeds therefore greatly reduces the risk that annealed and control fibrils will have different morphologies. To confirm that the annealing procedure does not cause mor- phological changes, we performed high-resolution cryo-electron microscopy on both annealed and control fibrils. By design, the high- resolution algorithms can only detect regular, periodic structural fea- tures of amyloid fibrils. No matter how large a sample is analysed, irregular, nonperiodic defects can only ever appear as noise to the algorithms, and are therefore averaged away during image processing53. We observed no significant differences in the raw images (Fig. 4a, b), and collated 2D class averages were indistinguishable between conditions (Fig. 4c-f). We finally collected statistics on these classes, and found that the proportions of twisted vs straight fibrils were almost identical in each sample (43:57 in annealed fibrils and 40:60 in control fibrils). Furthermore, the average crossover length was >300 nm in both samples, and the fibril widths matched within error. In other words, there are no detectable morphological differ- ences between annealed and control fibrils. The only possible major differences between the annealed and controlfibrils are thereforefibril length, and defect frequency and structure. Fig. 3 | Controlling fibril growth defect frequency by varying supersaturation levels. aDefects (purple) are initially formed by incorrect or incomplete binding of an incoming monomer onto the fibril end and subsequently kinetically trapped by attachment of further monomer layers on top of themisalignedmonomer. Defects have been illustrated as lateral dislocations of the entire cross-section for con- venience only: other defect structures are also possible. b at low enoughmonomer concentrations the rate of correct elongation and dissociation are close to equili- brium (the solubility limit). In this regime defect removal is faster than formation, and kinetic trapping is largely eliminated. c Aβ40 solubility decreases substantially with temperature from 64 to 37 °C ensuring elongation remains close to equili- brium at all times. Data are adapted from ref. 49, except for the solubility at 60 °C, which we measured by NMR as shown in Supplementary Fig. 3. d We grow fibrils designed to be largely free of kinetically trapped growth defects, by slowly redu- cing the temperature of a solution of initially slightly supersaturated monomeric Aβ40 from 60 to 37 °C by 1 °C per hour. This ensures production of significant quantities of fibrils without requiring supersaturation. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 5 www.nature.com/naturecommunications Despite their invisibility to the high-resolution cryo-EM algo- rithms, the existence of growth defects in Aβ40 and Aβ42 fibrils can nonetheless be confirmed directly from the raw cryo-EM images. This is because certain types of growthdefects producenonlocal signatures visible in raw micrographs. These include defects that punctuate transitions from one fibril morphology to another possessing a dif- ferent twist length (Fig. 4h-j), and branching sites (Fig. 4k). However, their quantification using the raw images is impractical due to their rarity combined with most fibrils being located in large, un-analysable clumps and tangles on the TEM grid51. More importantly, this count wouldmiss all growthdefects that affect only the adjacent fibril planes, such as dislocations and partiallymisfoldedmonomers, and that could conceivably greatly outnumber visible defects. Annealed fibrils possess far fewer secondary nucleation sites than normal fibrils Since Brichos also strongly inhibits Aβ40 aggregation at very low stoichiometries37, it is likely that, as was shown for Aβ42 fibrils, Brichos binds almost exclusively to secondary nucleation sites and not else- where along Aβ40 fibrils. We therefore again sought to quantify the number of secondary nucleation sites for both annealed and control fibrils by measuring the Brichos binding stoichiometry to Aβ40 fibrils. Fig. 4 | Cryo-EM data processing and statistics of control fibrils vs annealed fibrils. a, b Motion corrected micrographs showing control fibrils and annealed fibrils, respectively. c, d 2D class averages for twisted fibrils in control fibrils and annealed fibrils datasets, respectively, and measurements of fibril diameter at points along the fibril crossover. e, f 2D class averages for untwisted fibrils in 37 °C fibrils and annealed fibrils datasets, respectively. g summary of fibril morph population and morphology statistics. Cryo-TEM images of Aβ40 (h) or Aβ42 (i, j and k) fibrils showing evidence of growth defects, as pointed by the white arrows. The scale bar shown as white line is 50nm. The Aβ40 fibrils in (h) are formed in a solution of 10 μMAβ40 at pH 7.4, 37 °C, while Aβ42 fibrils in (i, j and k) are formed in a solution of 5 μM Aβ42 at pH 8, 37 °C. White arrows point to the locations of the defects. Further details of the cryo-EM processing workflow are found in Supplementary Fig. 7. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 6 www.nature.com/naturecommunications This time, we used fluorescence correlation spectroscopy (FCS) in addition to MDS to eliminate the possibility that differing Brichos binding stoichiometries for annealed and control fibrils are an artefact of either experimental technique. In both FCS and MDS experiments, we mixed different con- centrations of Alexa488-Brichos with fixed concentrations of non- fluorescent Aβ40fibrils and subsequentlymeasured the freeAlexa488- Brichos concentration. In FCS this was done by positioning the con- focal volume at a height where only diffusible Alexa488-Brichos can be detected, excluding those attached to the sedimented fibrils at the bottom of the dish (see Methods “Fluorescence correlation spectro- scopy”). In MDS, this was done by centrifuging the samples to remove fibril-bound Brichos and performing measurements on the super- natant (see Methods “Microfluidic Diffusional Sizing”). Visual inspec- tion of the resultant binding curves of free vs total Brichos (Fig. 5a, b) reveals annealed seeds to have a lower Brichos saturation concentra- tion compared to control seeds, indicating a lower secondary nuclea- tion site frequency, and supporting the hypothesis that they are growth defects. We then globally fit these two binding curves using a simple protein-ligand binding model54(Eq. 5). We found one Brichos binding site per 99monomers in control fibrils. This, alongside the fact that an additional and distinct site quantification technique was used com- pared to Aβ42, provides even more support for the rarity of Aβ sec- ondary nucleation sites. In annealed fibrils, the stoichiometry of these sites was reduced to one per 834 monomers. This implies that the number of secondary nucleation sites is reduced by approximately 88% by the annealing treatment relative to fibrils formed entirely at 37 °C. This large reduction validates a key prediction of the hypothesis that the secondary nucleation sites are growth defects. Note that, if Brichos also binds elsewhere along Aβ40 fibrils, the secondary nucleation site stoichiometry would be even lower than what was inferred here from the Brichos binding stoichiometry. Since the annealed and control fibril morphologies have been demonstrated to be the same on average, this would in turn imply an even greater reduction in secondary nucleation site incidence in annealed fibrils compared to control fibrils if they are indeed growth defects. Tri- merization of Brichos in solution38 could scale the measured free Bri- chos concentrations by a constant factor, but this will not affect the fitted binding stoichiometry, instead only changing the y-axis pro- portionality constants in our binding model. Annealed fibrils display considerably reduced secondary nucleation seeding efficiency compared to normal fibrils We next used 0.1 or 2μM monomer-equivalent of the second gen- eration fibrils (control seeds or annealed seeds) as preformed seeds in new aggregation reactions of 5μMmonomeric Aβ40 at 37 °C. We find that even though both control and annealed seeds accelerate the aggregation (Supplementary Fig. 1), the annealed seeds havemuch less seeding efficiency compared to control seeds in both lightly and heavily seeded conditions, Fig. 5c, d. The decrease in seeding efficiency of annealed seeds compared to control seeds could, in principle, reflect annealed fibrils having greater length, and thus possessing fewer elongation-competent fibril ends per unitfibrilmass, although such length differences should be limited by the iterative seeding approach used in their production. To test this possibility, aggregation data were fitted using the standard kinetic model for Aβ40 aggregation16, allowing the seed fibril length to differ between control seeds and annealed seeds (by allowing the seed fibril number concentration P0 to differ between conditions at fixed seed mass concentrationM0). Thismodel was unable to globally fit the data, even assuming extreme (≫trillion-fold) length differences, yielding pronounced misfits particularly at 2μMmonomer-equivalent of seeds (Fig. 5c). In fact, lower-resolution cryo-TEM images show no sign that annealed and control fibrils differ significantly in length (Supplementary Fig. 5). More realistic upper bounds of their average length difference (e.g. 10× or 100×) lead to markedly worse misfits (Supplementary Fig. 5), further strengthening the conclusion that length differences alone cannot explain the large difference in seeding propensity of these fibrils. Note that directly eliminating seed length differences by sonication was not feasible, because Aβ40 fibrils gen- erated under standard in vitro conditions are highly resistant to sonication51. Moreover, vigorous sonication of annealed and control fibrils in weaker fibril types such as Aβ4251 may never reduce them to the same length if fragmentation occurs preferentially at defect sites (see Supplementary Discussion). The reduced number of secondary nucleation sites in annealed fibrils canbe capturedby a kineticmodel that is identical in all respects to the standard model for Aβ40 aggregation16, but where the seed mass concentrationM0 ismultiplied byfitting parameter r, the fraction of remaining secondary nucleation sites compared to control fibrils (See Methods Sec. “Kinetic modelling” for modified kinetic model equations). We find that such a model, where effectively M0 can vary but not P0, can fit the measured kinetic data very well (Fig. 5d), using a value for r of 0.08, which is consistent with a 92% reduction in sec- ondary nucleation site numbers. This is remarkably consistent with the 88% reduction estimated using Brichos stoichiometry and is despite thismodel havingnomore degrees of freedom than the differing-seed- length model. In general, one might expect the ensemble of defect structures in annealed and control fibrils to differ somewhat. However, this close correspondence between secondary nucleation site stoi- chiometry and secondary nucleation propensity implies that there is little structural difference on average between the secondary nuclea- tion sites in annealed and control Aβ40 fibrils. Aβ40 secondary nucleation occurs predominantly at defects regardless of the supersaturation level Brichos can increase the Aβ40 aggregation half-time by at least 10-fold under the conditions used to generate the control fibrils37, which, given Aβ aggregation half times scale inversely with the cube root of the secondary nucleation rate55,56, implies a ≥1000-fold reduction in total secondary nucleation. As demonstrated above, Brichos effects this inhibition by blocking defect sites, so its causing this ≥1000-fold reduction requires that nucleation from non-defect surface sites on control fibrils must be at least 1000-fold slower than nucleation at defects in the absence of Brichos. Annealed Aβ40 fibrils grown at near- zero supersaturation still possess growth defects at Boltzmann distribution-governed stoichiometry, and their secondary nucleation rates and defect abundance are reduced by 8- to 12-fold relative to control fibrils. Consequently, even in supersaturation-free Aβ40 fibrils, secondary nucleation at defects remains at least about 50-fold faster than nucleation from other, non-defect surface sites under the same conditions. At equilibrium, the fraction of defective fibril planes vs total fibril planes peq is mathematically related to the free energy penalty ΔGdef for forming the least unstable possible types of defect by −RTln(peq) = ΔGdef (assuming defective layers are rare, see Supple- mentary Methods Sec. 3.1). peq is related to the equilibrium defect site stoichiometry per monomer seq by seq = ndef peq/x, where x is the number ofmonomers in the fibril cross-section and ndef the number of discrete defect sites per defective layer (Supplementary Methods Sec. 3.1). For instance, ndef = 2 in dislocation-type defects because dislocations expose fibril core on both sides of the fibril. In Supple- mentary Methods Sec. 3.1, we use these numbers to estimate the equilibrium defect-free energy penalty for several plausible values of x and ndef, finding ΔGdef to range from −14.1 to −17.8 kJ/mol. This being manymultiples of the thermal fluctuation energy RT, these defects are thus very thermodynamically unstable, as expected given their low stoichiometry. ΔGdef is also proportionally significant: approximately half of the per-monomer free energy for correct elongation (calculated Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 7 www.nature.com/naturecommunications in SupplementaryMethods Sec. 3.1 as − 32.3 kJ/mol at 45 °C). Note this calculation applies to the majority of defect structures seen in annealedfibrils. However, as shownabove, there is evidence to suggest that the defect structures in annealed and control fibrils are similar. If ΔGdef for Aβ42 fibrils is around half of the per-monomer free energy for correct elongation, as it is for Aβ40 fibrils, then switching to non-supersaturated growth conditions can be calculated to again reduce growth defect frequency by around 20-fold. Nonetheless, Bri- chos binding can also increase Aβ42 aggregation half-times by at least 6.5-fold46, corresponding to a >250-fold reduction in secondary nucleation rate. So, fibril growth defects should remain the dominant source of both Aβ40 and likely also Aβ42 secondary nucleation even without any supersaturation. (Note, older and less detailed kinetic models predict that the aggregation half times scale inversely with the square root of the secondary nucleation rate instead of the cube root10,16. Even if this were true, however, our conclusion would remain unchanged.) Discussion There are several reasons to expect that defects may act as secondary nucleation sites across many or even most types of amyloid fibrils. Diverse amyloid proteins form fibrils with highly varied surface che- mistries. About the only structural feature these fibrils have in com- mon is their hydrogen bonding between adjacent planes in the filaments, with significant differences in the packing of hydrophobic side-chains in the fibril core. The almost universal occurrence of rapid secondary nucleation across amyloid-forming proteins9,57 thus already suggests that secondary nucleation may have more to do with the Fig. 5 | Annealed fibrils have fewer secondary nucleation sites and lower sec- ondary nucleation rate than control fibrils. a and b Solutions of 4.5 μM (a) or 5 μM (b) Aβ40 fibrils (annealed or control) with 3 to 100nM Alexa488-Brichos were monitored by FCS (a) or MDS (b) to see the number of diffusing Alexa488-Brichos. Global fitting of both FCS and MDS data with a standard ligand-protein 1: 1 binding model (Eq. (5)) reveals the number of Alexa488-Brichos binding sites per Aβ40 molecule in the fibril for annealed seeds and for control seeds is s=1/99 (95% Confidence Interval = [1/116, 1/89]) and s=1/834 (95% Confidence Interval = [1/1216, 1/603]), respectively. Solid lines show the bootstrapmedian fit. The shaded regions represent bootstrap confidence bands, which illustrate the uncertainty in the model predictions across resampled datasets, as detailed in Methods Sec. “Fitting of ligand-protein binding model”. Sample sizes consisted of 1–3 repeats per condition. c,dThe aggregationof 5 μMAβwith 0.1 or 2μMmonomer-equivalent of annealed or control seedswasmonitored by ThT fluorescence in a plate reader and globally fitted to kineticmodels (lines). Themodel in (c) allowed annealed seeds to differ from control seeds only by their length and gave misfits with even a 100 trillion-fold average length difference. Using more plausible average fibril seed length differences (10–1000x) leads to even worse misfits (Supplementary Fig. 5). Sample sizes consisted of 2–3 repeats per condition. The model in d allowed annealed seeds to differ from control seeds only by their secondary nucleation site stoichiometry (see Methods Sec. “Kinetic modelling”) and gave good fits. Again, defects have been illustrated schematically as lateral dislocations for con- venience only. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 8 www.nature.com/naturecommunications amyloid core than with the fibril surface. Indeed, it has not yet been found to be possible to design an Aβ42 mutant that eliminates sec- ondary nucleation without also eliminating fibril growth11. Moreover, substitutions outside the amyloid fibril core do not strongly affect the ability of mutant fibrils to catalyse secondary nucleation of WT Aβ4258,59. Another structural feature shared by all amyloid fibrils is the presence of growth defects, which is guaranteed by the laws of ther- modynamics (the Boltzmann distribution) regardless of their exact structure. Since these defects can provide partial access to the fibril cross-section, they are a very strong candidate for secondary nuclea- tion sites. Such defects can provide a partial scaffold to surmount the entropic penalty associated with forming a new oligomer or fibril60–62. Additional support for this proposal is found beyond the field of amyloid fibril self-assembly: analogous defect-mediated secondary nucleation is well documented in metal and polymer crystallisation63–65. Brichos potently inhibits secondary nucleation by binding to fibril surfaces across multiple amyloid systems, including α-synuclein66, IAPP67, Aβ40, andAβ42. Inbothα-synuclein and IAPP, inhibition causes a greater than tenfold increase in aggregation half-time at a Brichos:- monomer ratio of around 1:10. These measurements imply that only a small fraction of fibril surface area—likely rare structural defects— drives secondary nucleation under these conditions, as established above for Aβ fibrils. The ability of Brichos to bind tightly and block such sites across diverse amyloids suggests that these nucleation- prone regions share structural features. Since amyloid surface sequences and properties vary widely, these shared features plausibly involve partial exposure of the cross-β core. If so, and if growth defects universally promote secondary nucleation, then Brichos may be an even more general secondary nucleation inhibitor than is currently appreciated. Possession of a universal affinity for cross-β structure would also explain why Brichos often displays at least some ability to inhibit elongation by binding fibril ends66,68, which share structural similarities with defect sites. A surface need not sharemany structural featureswith aggregates to promote their nucleation. Indeed, primary nucleation of fibrils (and unrelated aggregates) almost always happens on reaction vessel sur- faces in vitro, not in bulk57,69,70. By extension, other areas of the fibril surface may also act as secondary nucleation sites in some systems, in addition to growth defects. Regardless of whether significant off- defect nucleation occurs, multiple binding sites with diverse affinities likely exist on typical fibril surfaces. For example, binding studies on both Aβ42 fibrils and α-synuclein fibrils at physiological temperature and pH have revealed two classes of Brichos interaction: a rare, high- affinity site with nanomolarKD, and amuchweakermicromolar-KD site that may represent nonspecific surface binding24,66,71. In both systems, however, the high potency of secondary nucleation inhibition by Brichos21,66 implies that secondary nucleation occurs only at the tight site (see Supplementary Table 3 for KD comparison table). These results suggest a conceptual division between defect-associated nucleation (tight sites) and more generic surface adsorption (loose sites). Conversely, off-defect secondary nucleation along thewhole fibril has been proposed for α-synuclein under acidic conditions38. While substoichiometric inhibition by Brichos in that case was proposed to be caused by a large diffusion-driven effective volume for Brichos reliant on relatively weak Brichos binding affinity, the experimental observations are also consistent with a defect-based pathway. Impor- tantly, binding studies that assume a single class of sites can yield apparent average affinities or fail to detect rare high-affinity sites altogether. Kinetic modelling, therefore, provides a crucial com- plementary approach to determine inhibitor affinities at nucleation- active sites. The relationship between fibril morphology and growth defects merits further study. Thefibril core structureexposed at a defect could potentially be templated onto the newly forming fibril during sec- ondary nucleation, offering a possible molecular pathway for strain propagation. A recent investigation of Brichos binding to Aβ42 fibrils24 found that full inhibition of secondary nucleation by Brichos gives rise to a different, thinner morphology with a single filament consisting of two monomers per plane, as opposed to the more usual two-filament morph. This could indicate that secondary nucleation of Aβ42 propa- gates structure in some way. Alternatively, it could point to a key role of growth defects such as dislocations in anchoring the higher-order assembly of filaments. After secondary nucleation, the most notable secondary processes that can accelerate protein aggregation are fibril branching and fibril fragmentation. It is plausible that fragmentation is made possible by growth defects, which provide mechanically weak loci in fibrils. If so, and considering that rare fibril branching is also a defect-enabled phenomenon, the formation of growth defects could be the key molecular event driving autocatalytic amplification in all types offibril formation reactions. In the SupplementaryDiscussionwe suggest potential future experiments for testing this hypothesised relationship between defects and fragmentation, and explain why fragmentation cannot be a significant sink for defectswithout vigorous sonication. Crucially, the finding that secondary nucleation sites are defects could make structure-based drug design approaches possible for developing Alzheimer’s and Parkinson’s disease therapeutics. Sec- ondary nucleation sites on fibrils have long been recognised as a promising target for rational design of therapeutics21, being both the major source of toxic amyloid-β oligomers55, and obligate for rapid fibril proliferation9. Indeed, much of the interest in Brichos itself stems from its ability to bind to and block these sites to reduce the rate of secondary nucleation38. It also implies that reducing supersaturation levels in the brain (by e.g. reducing amyloidogenic proteinproduction) could be a particularly effective alternative therapeutic approach that not only directly lowers aggregation rates but also reduces defect stoichiometry, leading in turn to the formation of fewer toxic oligo- mers. Aβ40 solubility in vivo is expected to differ substantially from in vitro measurements because chemical compositions there are very different to standard in vitro buffers. Clearly, physiological con- centrations of Aβ40 must be above the local solubility limit for Aβ40 fibrils to accumulate in the brain. Since the central conclusion—that secondary nucleation predominantly occurs at growth defects—does not rely on supersaturation, it remains compatible with physiological settings even where local concentrations approach the solubility. In summary, our study shows that amyloid-β fibrils contain a small number of secondary nucleation sites, which can be further reduced by decreasing the level of supersaturation. These nucleation sites are identified as defects formed during fibril growth. The free energy penalty for a growth defect is roughly 50% of the per-monomer elon- gation free energy. A deeper understanding of defect-driven second- ary nucleation could open new opportunities for targeting amyloid formation in various diseases. By inhibiting secondary nucleation or reducing fibril defects, it may be possible to lower the formation of toxic species and slow disease progression. These findings offer a broaderperspectiveon themolecularmechanismsgoverning amyloid- related diseases andoffer a potential starting point for structure-based drug design for inhibiting amyloid secondary nucleation. Method Independent repeats in theplots correspond to separatemeasurement acquisitions performed on the same prepared sample, rather than measurements on independently prepared samples. Chemicals and consumables All Aβ42 or Aβ40 related experiments were carried out in 20mM sodium phosphate buffer (sodium dihydrogen phosphate dihydrate and di-sodium hydrogen phosphate) at pH 8.0 or pH 7.4, respectively. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 9 www.nature.com/naturecommunications The buffer componentswere purchased fromSigma-Aldrich (St. Louis, MO, USA), and all buffer solutions were prepared using Milli-Q water. The buffer was filtered through wwPTFE 0.2μm 50 mm disc filters (Fisher Scientific, Pittsburgh, PA, USA) and degassed prior to use. The dye Thioflavin T (ThT) formonitoring amyloid formationwasobtained from Calbiochem. Axygen® MaxyClear Snaplock Microtubes and CorningⓇ 96-well Half Area Black/Clear Flat Bottom PEGylated Poly- styrene Microplates (3881) were used as containers for protein solu- tions to minimise protein adsorption onto container surfaces. Expression and purification of Aβ42, Aβ40 and Brichos Recombinant peptide Aβ (M1-42) (MDAEFRHDSGYEVHHQKLVF- FAEDVGSNKGAIIGLMVGGVVIA), referred to here as Aβ42, Aβ(1- 40) (DAEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVV), refer- red to here as Aβ40, and Aβ(M1-40) (MDAEFRHDSGYEVHHQKLVF- FAEDVGSNKGAIIGLMVGGVV) were expressed as inclusion bodies in E. coli BL21(DE3)pLysS Star and purified using denaturation, anion- exchange chromatography, and size-exclusion chromatography to obtain highly pure monomeric peptide72,73. Aβ(1-40) was expressed in fusion with the EDDIE tag to enable a free Asp1, and its purification required additional steps for refolding/autocleavage of the EDDIE tag73. TheAβ(1-40) peptidewas used in all Aβ40experiments, except in themeasurements of solubility usingNMRspectroscopy, inwhich case Aβ(M1-40) was used. The proSP-C BRICHOS domain was expressed in E. coli BL21(DE3), isolated under denaturing conditions by Ni2+ affinity chromatography, and further purified following thrombin cleavage and ion-exchange chromatography37. Purified Aβ peptides were ali- quoted and stored at −80 °C. Purified BRICHOS was aliquoted and stored at −80 °C until use. Protein concentrations were determined by UV absorbance. Immediately prior to each experiment, Aβ42 or Aβ40 powder was dissolved in 6M guanidine hydrochloride (pH 8.5), and monomers were isolated by size exclusion chromatography (Superdex 75 column) in 20mM sodium phosphate buffer as described in chemicals and consumables section. ProSP-C Brichos domain was covalently labelled with the Alexa- 488 dye (Thermo Fisher Scientific, Waltham, US) by incubating the protein for three hours in 20mM sodium phosphate buffer, pH 8.0, at room temperatureunder gentle agitationwith one equivalent of Alexa- 488 carboxylic acid succinimidyl ester to label covalently primary amines. Following the labelling reaction, excess dye was removed by desalting on a G25 gel filtration column, followed by purification by anion exchange to remove any protein with multiple labels, and an additional desalting step on aG25 gel filtration columnbefore use. The labelling efficiency of Alexa-488 on Brichos is around 90% as indicated by relative absorbance measurements at 280 and 490nm, see Sup- plementary Fig. 6. Stated Brichos concentrations always refer to both labelled and unlabelled chaperone monomers together. Microfluidic diffusional sizing The fabrication and operation of the microfluidic diffusion device followed established procedures39,40. Briefly, the microfluidic chips were fabricated by using standard soft lithography techniques into polydimethylsiloxane74. The analyte sample and the buffer were introduced in the device through reservoirs connected to the inlets. The flow rates in the channels were controlled by applying a negative pressure at the single outlet channel by means of a syringe pump (neMESYS, Cetoni GmbH, Korbussen, Germany). Typical flow rates were in the range 60μl/h to 90μl/h and lateral diffusion profiles were recorded at twelve different positions (3.52, 5.29, 8.57, 10.33, 18.61, 20.37, 28.65, 30.41, 58.69, 60.45, 88.73 and 90.5mm) by standard epifluorescence microscopy using a cooled CCD camera (Evolve 512, Photometrics, Tucson, AZ, USA). The diffusion profiles were fitted by model simulations based on advection-diffusion equations under the assumption that the diffusivities of the species in the mixture are distributed according to a bimodal Gaussian distribution39. In this approach, the fitting parameters are the mean and the standard deviation of at least two size distributions, corresponding to the populations of the free chaperone (small size range) and of the cha- perone bound to the fibrils (large size range). In order to evaluate the binding between Brichos and Aβ42 fibrils during the aggregation process, samples with Alexa-488 labelled Bri- chos at concentrations of0.25, 0.35 or 0.50μMweremixedwith 24μM Aβ42 monomers and 20μM ThT and incubated in a 96-well plate in a Fluostar Optima plate reader (BMG Labtech, Ortenberg, Germany) at 37 °C with double orbital rotation (400 rpm) in order to generate sufficiently short fibrils for injecting into the microfluidic channels10. The aggregation kinetics were monitored by following the associated increase in ThT fluorescence. After completion of the reaction, the samples were incubated for at least 2 days at 21 °C to ensure equili- brium conditions. They were subsequently analysed by microfluidic diffusion sizing at 21 °C. To evaluate the binding between Brichos and mature Aβ42 fibrils, the same Aβ42 aggregation reaction was performed at 37 °C but without Brichos. After completion of the reaction, aliquots of Alexa-488 labelled Brichos in the concentration range between 0 and 1.5 μM were added to samples of the mature Aβ42 fibrils and incubated for at least 2 days at 21 °C to ensure equilibrium condi- tions. They were subsequently analysed by microfluidic diffusion sizing at 21 °C. To better constrain the Langmuir model during subsequent fitting, more measurements were taken at lower total Brichos concentrations, where the bound concentration changes more rapidly. Brichos binding to Aβ40 fibrils was evaluated by incubating 5μM Aβ40 fibrils (control or annealed) with Alexa-488 labelled Brichos at 12.5, 25, 50 or 100 nM for 10min at 37 °C. Before loading the samples into MDS, Aβ40 samples were centrifuged and only supernatant was injected, as Aβ40 fibrils are shake-resistant and remain too long in the microfluidic channels51. Samples were centrifuged at 17,850 rpm (~29,000 g) for 15min at 37 °C, and 4μL of each supernatant was loa- ded into microfluidic channels for analysis on a Fluidity One-M Serum instrument (Fluidics Inc., Cambridge, UK). The intensity of Alexa-488 was measured, and the artefactual contribution from free dye was removed using the free dye fractions determined in the FCS experiments. Kinetic experiments The kinetic assay for Aβ40 displayed in Fig. 5b, c starts with 5 μM monomer in 20mM pH 7.4 phosphate buffer with 10 μM ThT, with or without 0.1 or 2 μM monomer-equivalent of seeds. The seeds were generated as shown in Supplementary Fig. 4. After mixing the monomer and the seeds, the solutions were immediately incubated in a 96-well half-area clear flat-bottom PEGylated polystyrene microplate at 37 °C. The fluorescence was recorded in a Fluostar Omega plate reader (BMG Labtech) using a 440 nm excitation filter and a 480 nm emission filter, and used to monitor the extent of aggregation. The waiting time between the reading cycles was set to 0, meaning that the wells were continuously monitored throughout the kinetic measurements. This setup is designed to control the agitation frequency caused by reading wells consistently across all experiments46. The kinetic assays for Aβ42 were performed by incubating a solution of 3μM Aβ42 peptide, 1% seeds, and 20 μM ThT with or without different concentrations of Brichos in a non-binding 96 well plate in a plate reader Fluostar Optima (BMG Labtech, Ortenberg, Germany) at 37 °Cwith double orbital rotation (400 rpm). The validity of ThT fluorescence as a quantitative reporter of Aβ40 and Aβ42 fibril formation using this protocol has been established in multiple mechanistic studies10,16,75. Its use here complies with community requirements on chemical probes. Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 10 www.nature.com/naturecommunications Kinetic modelling Aβ42 aggregation. We describe Aβ42 aggregation using a chemical kinetics framework that accounts for the various microscopic steps of aggregation (Fig. 1a)16. The unperturbed kinetic equations for the aggregate number concentration P(t) and aggregate mass concentra- tion M(t) are: dP dt = k1m n1 + k2 m n2M, ð1aÞ dM dt =2k +mP = � dm dt , ð1bÞ wherem(t) is the monomer concentration, k1, k2, k+ are the rate con- stants for primary nucleation, secondary nucleation and elongation, and n1, n2 are the reaction orders of primary and secondary nucleation with respect to the monomer concentration. Fits of the aggregation curves in the presence of Brichos in Fig. 2d were performed using the Amylofit platform (which imple- ments analytical solutions to the kinetic equations (1a), (1b))43. The presence of Brichos was captured by means of an effective rate constant for secondary nucleation k2 that depends on chaperone concentration. Aβ40 aggregation. Equation (1) apply unmodified to Aβ40 aggre- gation seeded with control fibrils. When an Aβ40 aggregation reaction is seeded with annealed fibrils, Eqs. (1) still apply, but with modified initial condition M(0) = rM0, where M0 is the mass con- centration of seed fibrils and r is the fractional reduction in sec- ondary nucleation sites caused by the annealing process. The analytical solution to such equations is then identical to that used in Amylofit43, but with all instances of M(0) replaced by rM0. We used an offline version of Amylofit to globally fit the modified and unmodified analytical solutions to the kinetic data for unseeded reactions and for reactions seeded with annealed and control fibrils, and to determine r. Full fits are in Supplementary Fig. 1; Figure 5d omits the unseeded kinetic curve to provide greater resolution on the seeded kinetic curves. Misfits in Fig. 5c were performed using the standard kinetic model in the AmyloFit platform with different average seed fibril lengths allowed for the annealed and control fibrils. Solubility measurement The aggregation of 106, 106, 132, and 101 μMAβ (M1-40)monomers at 27, 45, 50, and 60 °C, respectively, was monitored by integrating the methyl group region of the Aβ (M1-40) 1H NMR spectra. The solubility was estimated from the relative integral of the methyl signals at the final plateau compared to the initial value. The concentration of Aβ (M1-40) at the initial value of NMR was determined using UV absor- bance. A detailed description of the solubility measurements is pro- vided in Supplementary method sec. 3.3, and the solubility of Aβ (M1- 40) monomers at 27, 45, 50, and 60 °C is provided in the Supple- mentary table 1. The solubility of Aβ (M1-40) at 60 °C is plotted as a single dot in Fig. 3c. NMR experiments were performed on a Bruker Avance III HD 900 spectrometer (Bruker Biospin, Rheinstetten, Ger- many), operating at a 1H resonance frequency of 899.8MHz and fitted with a 5mm cold probe. Fluorescence correlation spectroscopy Before FCSmeasurements, samples of 4500 nMAβ40fibrils and3–100 nM Alexa488-Brichos were incubated together in MatTek dishes (35mm, 10mm glass bottom, No. 1.5 glass) at 37 °C for 2 h in a desic- cator with water below to avoid evaporation. The measurements were conducted using a Zeiss 980 confocal laser scanningmicroscope (LSM 980, Carl Zeiss) using ZEN 3.10 software, equipped with a Zeiss C-Apochromat 40×−/1.2 NA water immersion objective. Excitation was performed at 488 nm, and fluorescence emission was collected within the 499-622 nm range. Each data point in Fig. 5a is based on 3–4 FCS measurements, 90 s measurements each. FCS data were acquired using Zeiss ZEN microscopy software (Zeiss). The free Brichos con- centration was determined by fitting the correlation curves at a time range of 10−5 to 1 s to theoretical models for diffusion of two species of differentmolecular weights, representing the diffusion of free dye and Brichos (Fig. SI2)76, using the model: Y = 1 N 1� f 2 1 + X τD1 � � ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 + X τD1S 2 q + f 2 1 + X τD2 � � ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 + X τD1S 2 q 2 64 3 75+ 1 whereN is the total number of particles, f is the fractional contribution of the component, X is correlation time, S is the structural parameter fixed to 5, τD1 and τD2 are the diffusion times of the two components, which are fixed to 27μs and 126μs, respectively, representing the diffusion time of free dye and Brichos. The resulting free Brichos particle number is shown in Supplementary Fig. 2c, and the average fraction of Brichos at the baseline or above the baseline (4%and 10%) is used to calculate the free Brichos concentration, which is fitted by the ligand-protein binding model for stoichiometry. Fitting of ligand-protein binding model We consider the 1:1 binding reaction: P + L"kon koff PL, where our ligand L is Brichos and our protein P is a secondary nucleation site. At binding equilibrium, the total ligand concentration cL is: cL = ½L�+ ½PL�= ½L�+ cP ½L� ½L�+KD , ð2Þ where [L] is the concentration of free Brichos, [PL] is the concentration of Brichos bound to secondary nucleation sites, cP is the total con- centration of secondary nucleation sites, and KD is the dissociation constant. This can be rearranged to: ½L�2 + ½L�ðKD + cP � cLÞ � cLKD =0: ð3Þ Finally: ½L�= � 0:5 � ðKD +M0s � cLÞ+ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0:25 � ðKD +M0s � cLÞ2 + cLKD q , ð4Þ where s is the number of secondary nucleation sites per monomer residue in the fibrils, and M0 is the fibril mass concentration. To determine the number of secondary nucleation sites per monomeric residue (s) in Fig. 5, the free Brichos concentration y as a function of total Brichos concentration x was fitted with the following ligand binding equation, which is a rearranged form of Eq. (4): y=a � ð�0:5B+ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0:25B2 + x � KD q Þ ð5aÞ B=KD + sM0 � x, ð5bÞ where a is an arbitrary proportionality constant, KD is the dissociation constant for Brichos binding, and the quantity in the brackets is the concentration of free Brichos. The concentration of Aβ40 fibrils is M0 = 4500 nM in FCS experiments, and M0 = 5000nM in MDS experiments. FCS and MDS datasets were fitted globally, sharing KD, s(annealed), and s(control), while allowing separate scaling factorsaFCS and aMDS for the two techniques. To assess parameter uncertainty, we used nonparametric strati- fied bootstrap resampling: one replicate per Brichos concentration was randomly drawn for each dataset, the global fit was repeated (Q = Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 11 www.nature.com/naturecommunications 1000 iterations), and the distributions of fitted parameters were summarised by medians and 95% confidence intervals. Confidence bands in theplotsweregenerated as the spreadof bootstrap-predicted curves. Cryo-EM sample preparation The samples were imaged using cryo-EM as described previously73. Specimens were plunged using an automated plunge freezer (Leica EMGP) set at 20 °C with 90% relative humidity. Thin liquid films were created on glow-discharged lacey carbon 300 mesh grids (Agar Sci- entific) and blotted with filter paper before being plunged into liquid ethane at −184 °C, ensuring vitrification in a glass-like state that prevents ice crystal formation and preserves the original micro- structures. The specimens were stored in liquid nitrogen until further use. Cryo-EM imaging The prepared specimen grids were transferred into the electron microscope (JEM 2200FS) using a Fischione model 2550 cryo-transfer tomography holder. Imagingwas performed at an acceleration voltage of 200 kV, with zero-loss images captured digitally using a TVIPS F416 camera and SerialEM 4.0.27 under low-dose conditions with a 10 eV energy filter. Image contrast was adjusted automatically using ImageJ version 1.53a. Cryo-EM data collection and processing The cryoEM datasets were collected at the SciLife lab node in Umeå, Sweden, using a Titan Krios electron microscope (Thermo Fisher) operated at 300 kVwith a Falcon4 detector and a Selectris energy filter (Thermo Fisher) was used operating with a 10 e−V slit. Datasets of size 8106movies and 7990movieswere collected for control and annealed fibrils, respectively, at a nominal magnification of 130,000× was set yielding a pixel size of 0.92 Å. A defocus range of −1.4 to −2.6μmand a total dose of 40 e−/Å2 were used over an exposure time of 4.39 seconds. The rawEERmovieswere fractionated, aligned and summedusing motion correction in RELION-477 and CTF estimation was done for micrographs using CTFFIND478. Manual picking of fibrils was done until roughly 100,000 particles of 300 pixel box size were picked, and the extracted segment was used to train a separate picking model for each dataset in Topaz, which is incorporated in RELION-4.077. Iterative rounds of 2D classification were performed to remove picking arte- facts, such as carbon edges and ice contamination, until only classes that could be protein fibrils remained. For 2D classification, a box size of 600 pixels was used. Twisted and straight fibrils were then sepa- rated out and a final round of 2D classification was performed sepa- rately for both sets of fibrils. Note that, due to current software limitations, the average crossover length is too large for 3D fibril structures to be calculated; however, in this study, these 3D structures are not needed since the 2D classes already rule out significant structural differences in the fibrils. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability Source data are provided with this paper. The raw data generated in this study have been uploaded to figshare [https://doi.org/10.6084/ m9.figshare.28716485]. Source data are provided with this paper. Code availability Simple Python code used to fit data in Figure 5 is available on CodeOcean [https://doi.org/10.24433/CO.3160717.v1]. References 1. Sawaya, M. R., Hughes, M. P., Rodriguez, J. A., Riek, R. & Eisenberg, D. S. The expanding amyloid family: structure, stability, function, and pathogenesis. Cell 184, 4857–4873 (2021). 2. Hardy, J. & Selkoe, D. J. The amyloid hypothesis of alzheimer’s disease: progress and problems on the road to therapeutics. sci- ence 297, 353–356 (2002). 3. Selkoe, D. &Hardy, J. The amyloid hypothesis of alzheimer’s disease at 25 years. EMBO Mol. Med. 8, 595–608 (2012). 4. Ittner, L. M. & Götz, J. Amyloid-β and tau—a toxic pas de deux in alzheimer’s disease. Nat. Rev. Neurosci. 12, 67–72 (2011). 5. Oosawa, F. & Kasai, M. A theory of linear and helical aggregations of macromolecules. J. Mol. Biol. 4, 10–21 (1962). 6. Ferrone, F. A., Hofrichter, J., Sunshine, H. R. & Eaton, W. A. Kinetic studies on photolysis-induced gelation of sickle cell hemoglobin suggest a new mechanism. Biophys. J. 32, 361–380 (1980). 7. Ferrone, F. A., Hofrichter, J. & Eaton, W. A. Kinetics of sickle hemoglobin polymerization. I. Studies using temperature-jump and laser photolysis techniques. J. Mol. Biol. 183, 591–610 (1985). 8. Ferrone, F. A., Hofrichter, J. & Eaton, W. A. Kinetics of sickle hemoglobin polymerization. ii. a double nucleation mechanism. J. Mol. Biol. 183, 611–631 (1985). 9. Meisl, G. et al. Uncovering the universality of self-replication in protein aggregation and its link to disease. Sci. Adv. from review, 6831 (2022). 10. Cohen, S. I. et al. Proliferation of amyloid-β42 aggregates occurs through a secondary nucleation mechanism. Proc. Natl. Acad. Sci. 110, 9758–9763 (2013). 11. Törnquist, M. et al. Secondary nucleation in amyloid formation. Chem. Commun. 54, 8667–8684 (2018). 12. Ruschak, A.M.&Miranker, A. D. Fiber-dependent amyloid formation as catalysis of an existing reaction pathway. Proc. Natl. Acad. Sci. 104, 12341–12346 (2007). 13. Schlamadinger, D. E. & Miranker, A. D. Fiber-dependent and-inde- pendent toxicity of islet amyloid polypeptide. Biophys. J. 107, 2559–2566 (2014). 14. Buell, A. K. et al. Solution conditions determine the relative impor- tance of nucleation and growth processes in α-synuclein aggre- gation. Proc. Natl. Acad. Sci. 111, 7671–7676 (2014). 15. Rodriguez Camargo, D. C. et al. Surface-catalyzed secondary nucleation dominates the generation of toxic iapp aggregates. Front. Mol. Biosci. 8, 757425 (2021). 16. Meisl, G. et al. Differences in nucleation behavior underlie the contrasting aggregation kinetics of the Aβ40 and Aβ42 peptides. Proc. Natl. Acad. Sci. USA 111, 9384–9389 (2014). 17. Dear, A. J. et al. Thecatalytic natureof protein aggregation. J. Chem. Phys. 152, 045101 (2020). 18. Miers, H. A. & Isaac, F. Xlvii.—the refractive indices of crystallising solutions,with especial reference to thepassage from themetastable to the labile condition. J. Chem. Soc. Trans. 89, 413–454 (1906). 19. Botsaris, G.D. Secondary nucleation—a review. Industrial Crystal- lization, 3–22 (1976). 20. Agrawal, S. & Paterson, A. Secondary nucleation: mechanisms and models. Chem. Eng. Commun. 202, 698–706 (2015). 21. Cohen, S. I. A. et al. A molecular chaperone breaks the catalytic cycle that generates toxic Aβ oligomers. Nat. Struct. Mol. Biol. 22, 207–213 (2015). 22. Biversål, H. et al. Dissociation of a brichos trimer into monomers leads to increased inhibitory effect on aβ42 fibril formation. Bio- chim. Biophys. Acta 1854, 835–843 (2015). 23. Pansieri, J. et al. Templating s100a9 amyloids on aβ fibrillar surfaces revealed by charge detection mass spectrometry, microscopy, kinetic and microfluidic analyses. Chem. Sci. 11, 7031–7039 (2020). Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 12 https://doi.org/10.6084/m9.figshare.28716485 https://doi.org/10.6084/m9.figshare.28716485 https://doi.org/10.24433/CO.3160717.v1 www.nature.com/naturecommunications 24. Kumar, R. et al. Identification of potential aggregation hotspots on aβ42 fibrils blocked by the anti-amyloid chaperone-like brichos domain. Nat. Commun. 15, 965 (2024). 25. Curk, S. et al. Self-replication of a β 42 aggregates occurs on small and isolated fibril sites. Proc. Natl. Acad. Sci. 121, 2220075121 (2024). 26. Zimmermann,M. R. et al.Mechanismof secondary nucleation at the single fibril level from direct observations of aβ42 aggregation. J. Am. Chem. Soc. 143, 16621–16629 (2021). 27. Nirmalraj, P. N., Bhattacharya, S. & Thompson, D. Accelerated alz- heimer’s aβ-42 secondary nucleation chronologically visualized on fibril surfaces. Sci. Adv. 10, 5059 (2024). 28. Huang, C., Ghanati, E. & Schmit, J. D. Theory of sequence effects in amyloid aggregation. J. Phys. Chem. B 122, 5567–5578 (2018). 29. Akram, S. A. et al. Competing addition processes give distinct growth regimes in the assembly of 1d filaments. Biophys. J. 124, 778–788 (2025). 30. Schmit, J. D. Kinetic theory of amyloid fibril templating. J. Chem. Phys. 138, 185102 (2013). 31. Larsen, J. A. et al. The mechanism of amyloid fibril growth from ϕ- value analysis. Nat. Chem. 17, 403–411 (2025). 32. McPherson, A. & Kuznetsov, Y. G. Mechanisms, kinetics, impurities and defects: consequences in macromolecular crystallization. Struct. Biol. Cryst. Commun. 70, 384–403 (2014). 33. Jarrett, J. T. & Lansbury, P. T. Seeding “one-dimensional crystal- lization" of amyloid: a pathogenic mechanism in Alzheimer’s dis- ease and scrapie? Cell 73, 1055–1058 (1993). 34. Hur, S.-M. et al. Molecular pathways for defect annihilation in direc- ted self-assembly. Proc. Natl. Acad. Sci. 112, 14144–14149 (2015). 35. Pauls, A., Dunnum, A., Martin, A., Huth, A., Du, C., Thuo, M. Cor- recting edge defects in self-assembled monolayers through ther- mal annealing. ChemPhysChem, 202400626 (2024). 36. Lawaczeck, R., Kainosho, M. & Chan, S. I. The formation and annealing of structural defects in lipid bilayer vesicles. Biochim. Biophys. Acta (BBA) - Biomembr. 443, 313–330 (1976). 37. Willander, H. et al. Brichos domains efficiently delay fibrillation of amyloid β-peptide. J. Biol. Chem. 287, 31608–31617 (2012). 38. Ghosh, D. et al. The inhibitory action of the chaperone brichos against the α-synuclein secondary nucleation pathway. Nat. Com- mun. 15, 10038 (2024). 39. Müller, T. et al. Particle-based monte-carlo simulations of steady- state mass transport at intermediate péclet numbers. Int. J. Non- linear Sci. Numer. Simul. 17, 175–183 (2016). 40. Arosio, P. et al. Microfluidic diffusion analysis of the sizes and interactions of proteins under native solution conditions. ACSNano 10, 333–341 (2016). 41. Wilkinson, K.D. Quantitative analysis of protein-protein interactions. In C. L.Meyerkord, H. Fu (Eds.) Protein-Protein Interactions:Methods and Applications, 15–31 (2004). 42. Willander, H. et al. High-resolution structure of a brichos domain and its implications for anti-amyloid chaperone activity on lung surfactant protein c. Proc. Natl. Acad. Sci. 109, 2325–2329 (2012). 43. Meisl, G. et al. Molecular mechanisms of protein aggregation from global fitting of kinetic models. Nat. Protoc. 11, 252–272 (2016). 44. Lattanzi, V. et al. Amyloid β 42 fibril structure based on small-angle scattering. Proc. Natl. Acad. Sci. 118, 2112783118 (2021). 45. Colvin, M. T. et al. Atomic resolution structure of monomorphic aβ42 amyloid fibrils. J. Am. Chem. Soc. 138, 9663–9674 (2016). 46. Axell, E. et al. The role of shear forces in primary and secondary nucleation of amyloid fibrils. Proc. Natl. Acad. Sci. 121, 2322572121 (2024). 47. Kheterpal, I., Zhou, S., Cook, K. D. & Wetzel, R. Aβ amyloid fibrils possess a core structure highly resistant to hydrogen exchange. Proc. Natl. Acad. Sci. 97, 13597–13601 (2000). 48. Lührs, T. et al. 3d structure of alzheimer’s amyloid-β(1–42) fibrils. Proc. Natl. Acad. Sci. 102, 17342–17347 (2005). 49. Hicks, D.E. Thermophysical properties of the Amyloid Beta protein from differential scanning calorimetry. Master’s thesis (Department of Chemical Engineering, University of Tennessee, 2005). 50. Lattanzi, V., Bernfur, K., Sparr, E., Olsson, U. & Linse, S. Solubility of aβ40 peptide. JCIS Open 4, 100024 (2021). 51. Dear, A. J., Meisl, G., Hu, J., Knowles, T. P. J. & Linse, S. Kinetics of seeded protein aggregation: Theory and application. J. Chem. Phys. 163, 045101 (2025). 52. Thacker, D. et al. The role of fibril structure and surface hydro- phobicity in secondary nucleation of amyloid fibrils. Proc. Natl. Acad. Sci. U.S.A. 117, 25272–25283 (2020). 53. Scheres, S.H.W. Processing of structurally heterogeneous cryo-em data in relion. In: Crowther, R.A. (ed.) The Resolution Revolution: Recent Advances in cryoEM. Methods in Enzymology, vol. 579, pp. 125–157 (Academic Press, 2016). 54. Rodbard, D., Feldman, H. Theory of protein-ligand interaction. In: Methods in Enzymology vol. 36, pp. 3–16 (Elsevier, 1975). 55. Michaels, T. C. T. et al. Dynamics of oligomer populations formed during the aggregation of alzheimer’s aβ42 peptide.Nat. Chem. 12, 445–451 (2020). 56. Michaels, T. C., Dear, A. J., Cohen, S., Vendruscolo, M. & Knowles, T. P. J. Kinetic profiling of therapeutic strategies for inhibiting the formation of amyloid oligomers. J. Chem. Phys. 156, 164904 (2022). 57. Dear, A.J., Meisl, G., Taylor, C.G., Palmiero, U.C., Stubbe, S.N., Liu, Q., Arosio, P., Linse, S., Knowles, T.P.J., Andreasen, M.: Surface Effects on Functional Amyloid Formation (Nanoscale, 2024). 58. Thacker, D., Willas, A., Dear, A. J. & Linse, S. Role of hydrophobicity at the n-terminal region of aβ42 in secondary nucleation. ACS Chem. Neurosci. 13, 3477–3487 (2022). 59. Sanagavarapu, K. et al. Serine phosphorylation mimics of aβ form distinct, non-cross-seeding fibril morphs. Chem. Sci. (2024). 60. Zhang, L. & Schmit, J. D. Pseudo-one-dimensional nucleation in dilute polymer solutions. Phys. Rev. E 93, 060401 (2016). 61. Zhang, L. & Schmit, J. D. Theory of amyloid fibril nucleation from folded proteins. Isr. J. Chem. 57, 738–749 (2017). 62. Phan, T. M. & Schmit, J. D. Conformational entropy limits the tran- sition from nucleation to elongation in amyloid aggregation. Bio- phys. J. 121, 2931–2939 (2022). 63. Sounart, T. L. et al. Secondary nucleation and growth of ZnO. J. Am. Chem. Soc. 129, 15786–15793 (2007). 64. Jiang, J. et al. Defect-induced secondary crystals drive two- dimensional to three-dimensional morphological evolution in the co-self-assembly of polyferrocenylsilane block copolymer and homopolymer. J. Am. Chem. Soc. 145, 28096–28110 (2023). 65. Majumder, S., Busch, H., Poudel, P., Mecking, S. & Reiter, G. Growth kinetics of stacks of lamellar polymer crystals.Macromolecules 51, 8738–8745 (2018). 66. Adam, L. et al. Specific inhibition of α-synuclein oligomer genera- tion and toxicity by the chaperone domain bri2 brichos. Protein Sci. 33, 5091 (2024). 67. Oskarsson, M. E. et al. Brichos domain of bri2 inhibits islet amyloid polypeptide (iapp) fibril formation and toxicity in human beta cells. Proc. Natl. Acad. Sci. 115, 2752–2761 (2018). 68. Arosio, P. et al. Kinetic analysis reveals the diversity of microscopic mechanisms through which molecular chaperones suppress amy- loid formation. Nat. Commun. 7, 10948 (2016). 69. Toprakcioglu, Z. et al. Adsorption free energy predicts amyloid protein nucleation rates. Proc. Natl. Acad. Sci. USA 119, 2109718119 (2022). 70. Grigolato, F. & Arosio, P. The role of surfaces on amyloid formation. Biophys. Chem. 270, 106533 (2021). Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 13 www.nature.com/naturecommunications 71. Leppert, A. et al. Smallest secondary nucleation competent aβ aggregates probed by an ATP-independent molecular chaperone domain. Biochemistry 60, 678–688 (2021). 72. Walsh, D.M. et al. A facilemethod for expression and purification of the alzheimer’s disease-associated amyloid β-peptide. FEBS J. 276, 1266–1281 (2009). 73. Hu, J., Linse, S. & Sparr, E. Ganglioside micelles affect amyloid β aggregation by coassembly. ACS Chem. Neurosci. 14, 4335–4343 (2023). 74. Qin, D., Xia, Y. & Whitesides, G. M. Soft lithography for micro-and nanoscale patterning. Nat. Protoc. 5, 491 (2010). 75. Cukalevski, R. et al. TheAβ40andAβ42peptides self-assemble into separate homomolecular fibrils in binary mixtures but cross-react during primary nucleation. Chem. Sci. 6, 4215–4233 (2015). 76. Rauer, B., Neumann, E., Widengren, J. & Rigler, R. Fluorescence correlation spectrometry of the interaction kinetics of tetra- methylrhodamin α-bungarotoxin with torpedo californica acet- ylcholine receptor. Biophys. Chem. 58, 3–12 (1996). 77. Kimanius, D., Dong, L., Sharov, G., Nakane, T. & Scheres, S. H. New tools for automated cryo-em single-particle analysis in relion-4.0. Biochem. J. 478, 4169–4185 (2021). 78. Rohou, A. & Grigorieff, N. Ctffind4: Fast and accurate defocus esti- mation fromelectronmicrographs. J. Struct. Biol. 192, 216–221 (2015). Acknowledgements This work was supported by the Swedish Research Council (2019-02397 to E.S., 2015-00143 to S.L., and 2022-06641 to S.L. and E.S.), and the GenerationNano project, the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 945378 (S.L. co-PI). We acknowledge support from the Wellcome Trust (T.P.J.K.), the CambridgeCentre for Misfolding Diseases (T.P.J.K.), the BBSRC (T.P.J.K.), the Frances and Augustus Newman Foundation (T.P.J.K.), the ERC PhysProt (agreement n 337969) (T.S., T.P.J.K., S.L.), ETC StG “NEPA” (A.Š. and S.C.), the Royal Society (S.C., A.S.), the ERASMUS Programme (T.S.), and The Danish Council for Independent Research ∣ Natural Sciences (FNU-11-113326) (M.A.). This work was also funded by the Novo Nordisk Foundation (#NNF19OC0054635 to S.L.), ETH Zürich (T.C.T.M.), and the Swiss National Science Foundation (grant no 219703 to A.J.D. and T.C.T.M.). We acknowledge the use of the nano-Characterisation and nano- Manufacturing Research Equipment (nCHREM) facility for access to microscopy instrumentation. We are grateful to the late Professor Sir Christopher Dobson for invaluable conversations regarding the micro- fluidic diffusional sizing experiments. We are also grateful to Quentin A. E. Peter and Thomas Müller for their guidance on microfluidic device design. The cuvette-filled icon in Fig. 3d is by Servier [https://smart. servier.com/]. It is licensed under CC-BY 3.0 Unported [https:// creativecommons.org/licenses/by/3.0/]. The authors would like to acknowledge Umeå Centre for Electron Microscopy (UCEM) for techni- cal assistance and access to electron microscopy. Support was pro- vided by SciLifeLab national Cryo-EM Unit at Umeå University. Author contributions Conceptualisation: A.J.D., T.P.J.K., S.L., E.Sparr. Data curation: J.H., T.S., D.T. Formal analysis: A.J.D., J.H., D.T., T.C.T.M. Funding acquisition: E.Sparr, S.L., T.P.J.K., T.C.T.M., A.Š. Investigation: J.H., T.S., D.T., S.L., E.A., E.Stemme,U.L., S.W., P.A., S.C.,M.A.Methodology: A.J.D., J.H., S.L., E.Sparr, T.C.T.M., T.S, J.D.S. Project administration: A.J.D., T.P.J.K., T.C.T.M., S.L., E.Sparr. Resources: S.L., E.Sparr, T.P.J.K., S.W. Software: J.H., T.S., D.T., G.M., T.C.T.M., A.J.D. Supervision: A.J.D., S.L., E.Sparr, T.P.J.K., T.C.T.M., P.A., A.Š. Validation: J.H., T.S., S.W. Visualisation: J.H., T.S., T.C.T.M., S.L., D.T. Writing - original draft: A.J.D., J.H., T.C.T.M., T.S. Writing - review & editing: J.H., T.S., D.T., E.A., E. Stemme, U.L., S.W., G.M., S.C., M.A., M.V., A.Š., J.D.S., T.P.J.K., E.Sparr, S.L., T.C.T.M., and A.J.D. Competing interests The authors declare no competing interests. Additional information Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-69377-1. Correspondence and requests for materials should be addressed to Alexander J. Dear. Peer review information Nature Communications thanks Devrim Kilinc, Christofer Lendel, and the other anonymous reviewer(s) for their con- tribution to the peer review of this work. A peer review file is available. Reprints and permissions information is available at http://www.nature.com/reprints Publisher’s note Springer Nature remains neutral with regard to jur- isdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. © The Author(s) 2026 1Division of Physical Chemistry, Department of Chemistry, Lund University, Lund, Sweden. 2Department of Biochemistry and Structural Biology, Lund University, Lund, Sweden. 3Centre for Misfolding Diseases, Yusuf Hamied Department of Chemistry, University of Cambridge, Cambridge, UK. 4Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg, Esch-sur-Alzette, Luxembourg. 5Astbury Centre for Structural Molecular Biology, School of Molecular and Cellular Biology, Faculty of Biological Sciences, University of Leeds, Leeds, UK. 6Living Systems Institute, University of Exeter, Exeter, Devon, UK. 7Department of Applied Physics, Biophysics Group, SciLifeLab, Royal Institute of Technology-KTH, Solna, Sweden. 8UK Dementia Research Institute, University of Cambridge, Cambridge, UK. 9Institute of Science and Technology Austria, Klosterneuburg, Austria. 10Department of Biomedicine, Aarhus University, Aarhus, Denmark. 11Department of Chemistry and Applied Biosciences, Institute for Chemical and Bioengineering, ETH Zurich, Zurich, Switzerland. 12Department of Physics, Kansas State University, Manhattan, USA. 13Cavendish Laboratory, University of Cambridge, Cambridge, UK. 14Institute of Biochemistry, Department of Biology, ETH Zurich, Zurich, Switzerland. 15Bringing Materials to Life Initiative, ETH Zurich, Zurich, Switzerland. e-mail: alexander.dear@bc.biol.ethz.ch Article https://doi.org/10.1038/s41467-026-69377-1 Nature Communications | (2026) 17:1933 14 https://smart.servier.com/ https://smart.servier.com/ https://creativecommons.org/licenses/by/3.0/ https://creativecommons.org/licenses/by/3.0/ https://doi.org/10.1038/s41467-026-69377-1 http://www.nature.com/reprints http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ mailto:alexander.dear@bc.biol.ethz.ch www.nature.com/naturecommunications Structural defects in amyloid-β fibrils drive secondary nucleation Results Brichos binds Aβ42 amyloid fibrils tightly with low stoichiometry Modulation of Aβ42 aggregation by Brichos reveals rare secondary nucleation sites Fibrils possessing fewer growth defects can be generated without measurably altering morphology Annealed fibrils possess far fewer secondary nucleation sites than normal fibrils Annealed fibrils display considerably reduced secondary nucleation seeding efficiency compared to normal fibrils Aβ40 secondary nucleation occurs predominantly at defects regardless of the supersaturation level Discussion Method Chemicals and consumables Expression and purification of Aβ42, Aβ40 and Brichos Microfluidic diffusional sizing Kinetic experiments Kinetic modelling Aβ42 aggregation Aβ40 aggregation Solubility measurement Fluorescence correlation spectroscopy Fitting of ligand-protein binding model Cryo-EM sample preparation Cryo-EM imaging Cryo-EM data collection and processing Reporting summary Data availability Code availability References Acknowledgements Author contributions Competing interests Additional information