21 Novembre 2026

Psicologia Punto One | Psicologia e Psicologi nel Web

Menu

Menu

SCHEDA SCIENTIFICA2026Altra ricerca

Two functionally distinct amyloid-β intermediates link aggregation pathways to pathology in Alzheimer's disease.

Agnieszka Urbanek, Alexander J Dear, David P Rhodes, Emma F Garland, Hollie E Wareing, Harry J Baker, Willem Molenkamp, Ruixin Zhang, Gefei Chen, Axel Abelein, Simon M Bell, Juan A Varela, Thomas C T Michaels, J Robin Highley, Georg Meisl, Jan Johansson, Suman De

Nature communications · 2026

Abstract originale

Neurotoxic intermediates formed during amyloid-β (Aβ) aggregation are thought to drive Alzheimer's disease, yet their rarity and transient nature have obscured their identity and formation pathways in the human brain. Here, we combine super-resolution microscopy and single-molecule functional imaging with kinetic modelling to classify Aβ intermediates according to their membrane-disrupting activity and connect their damaging effects to the aggregation pathways that generate them. Applying this framework to recombinant Aβ42, an Alzheimer's knock-in mouse model and human Alzheimer's brain tissue, we find two functionally distinct intermediate populations present across all systems. They form through separate aggregation routes and differ in their neurotoxicity, abundance and susceptibility to inhibition. Both damage neurons and glia, but only one converts into plaques. Suppressing these intermediates pharmacologically in mice reduces plaque burden, attenuates astrogliosis and improves cognition. Our findings connect model systems with human pathology and identify intermediate formation pathways as actionable therapeutic targets across protein-misfolding disorders.

Parole chiave

Alzheimer DiseaseAnimalsAmyloid beta-PeptidesHumansBrainMicePlaque, AmyloidNeuronsDisease Models, AnimalMice, TransgenicProtein Aggregation, PathologicalPeptide FragmentsNeurogliaFemale
Leggi il testo integrale originale

Testo estratto dalla pubblicazione: sono esclusi immagini, media, supplementi e parti soggette a diritti distinti.

Two functionally distinct amyloid-β intermediates link aggregation pathways to pathology in Alzheimer’s disease

Neurotoxic intermediates formed during amyloid-β (Aβ) aggregation are thought to drive Alzheimer’s disease, yet their rarity and transient nature have obscured their identity and formation pathways in the human brain. Here, we combine super-resolution microscopy and single-molecule functional imaging with kinetic modelling to classify Aβ intermediates according to their membrane-disrupting activity and connect their damaging effects to the aggregation pathways that generate them. Applying this framework to recombinant Aβ42, an Alzheimer’s knock-in mouse model and human Alzheimer’s brain tissue, we find two functionally distinct intermediate populations present across all systems. They form through separate aggregation routes and differ in their neurotoxicity, abundance and susceptibility to inhibition. Both damage neurons and glia, but only one converts into plaques. Suppressing these intermediates pharmacologically in mice reduces plaque burden, attenuates astrogliosis and improves cognition. Our findings connect model systems with human pathology and identify intermediate formation pathways as actionable therapeutic targets across protein-misfolding disorders.

Authors combine super-resolution microscopy and single-molecule functional imaging with kinetic modelling to classify Aβ intermediates according to their membrane-disrupting activity and connect their damaging effects to the aggregation pathways that generate them, connecting model systems with human pathology.

Introduction

Alzheimer’s disease (AD) is the most prevalent form of dementia, affecting over fifty million individuals worldwide. One of its defining pathological hallmarks is the extracellular deposition of amyloid-β (Aβ) peptides into plaques1, a process that begins with the misfolding and aggregation of Aβ monomers into soluble intermediates, which subsequently assemble into insoluble fibrils that accumulate as amyloid plaques2,3. As Aβ aggregates and deposits in the brain, levels of soluble Aβ in cerebrospinal fluid decline, providing one of the earliest detectable abnormalities in AD4. Yet the relevance of this aggregation pathway to disease progression remains contentious: Aβ plaque burden correlates poorly with cognitive decline5,6, and immunotherapies aimed at clearance of plaques have shown only modest clinical benefit7,8. These observations raise fundamental questions about the causal role of Aβ plaques in AD9,10. Mounting evidence instead implicates Aβ intermediates, such as oligomers, formed en route to plaques, as key drivers of neurodegeneration and better correlate with cognitive dysfunction11–16.

Among the Aβ isoforms, the 42-residue Aβ peptide, Aβ42, is particularly prone to aggregation17 and disproportionately enriched in amyloid plaques in AD brains18. It is also deposited into amyloid plaques earlier in the disease process compared to other Aβ isoforms19. As a result, the role of Aβ42 in AD is extensively studied in model systems, primarily through recombinant protein assays and transgenic animal models. Whilst these approaches have provided key insights into Aβ dysfunction and associated toxicity, they have certain limitations. Recombinant studies typically require Aβ concentrations several orders of magnitude higher than those found in the human brain20. Even the most advanced knock-in mouse models fail to fully recapitulate the spatial and temporal complexity of human AD pathology21,22. These inconsistencies, along with the limited success of plaque-targeting therapies, raise doubts about whether aggregation pathways observed in model systems truly parallel those in the human brain. They further cast doubt on whether intermediates detected in controlled assays are present in AD tissue and whether these are indeed the toxic species driving neuronal injury. Without a direct molecular and functional bridge linking model systems to human disease, the identity of the species that initiate the pathogenic cascade, and the pathways by which they form in the brain, have remained elusive.

Progress has been hindered by the fact that Aβ aggregation cannot be monitored directly in the living human brain. Histopathological analyses of postmortem tissue provide static snapshots, the gold standard for confirming the presence of pathology, but offer little insight into the underlying aggregation dynamics. Although structural methods such as optical imaging23,24, cryo-EM25,26, or NMR27,28 have detected Aβ aggregates in brain-derived extracts, they cannot link these structures to their functional consequences or to the pathways that give rise to them. Conversely, functional assays demonstrated the neurotoxicity of brain-derived Aβ assemblies11,29, yet they offer no information on how intermediate species emerge or evolve in the diseased brain. Consequently, current therapeutic strategies continue to emphasise late-stage plaque removal, long after irreversible neurodegeneration has occurred, reinforcing the prevailing notion that early intermediates remain effectively undruggable30.

To address this challenge, we developed a digital platform that detects individual Aβ species based on their damaging potential at single-aggregate resolution. This assay quantifies Aβ’s ability to disrupt membranes and trigger Ca2+ influx. The assay builds on extensive evidence that Aβ toxicity in AD involves disruption of neuronal membranes and Ca2+ dyshomeostasis, which in turn leads to synaptic failure, neural network dysfunction, and cognitive decline31–34. Aβ-induced Ca2+ dysregulation has been directly linked to memory deficits35, and Ca2+ alterations have been observed in individuals with mild cognitive impairment36. Yet the identity of the responsible species, and how to intercept their formation to prevent damage, remain unresolved. Our assay overcomes this limitation by sampling Aβ species formed along the in vitro aggregation pathway as well as in mouse models and human postmortem Alzheimer’s brains. It detects Aβ-induced membrane permeabilisation and Ca2+-ion influx in real time with high sensitivity, registering events even when triggered by the passage of a single Ca2+-ion. This enables classification of toxic Aβ species by the strength of their membrane-damaging activity.

Using this approach, we identified two functionally distinct Aβ aggregation intermediates present across recombinant peptides, Alzheimer’s mouse models, and human brain. By integrating functional data with kinetic theory, we mapped how these intermediates form, evolve, drive toxicity, and respond to therapeutic intervention. Our findings establish a link between models and human disease, connect neurotoxic potency to aggregation pathways, and provide a framework for selectively targeting the transient Aβ assemblies that drive neurodegeneration and AD progression.

Results

Functional fingerprinting of Aβ aggregates at the single-aggregate level

To quantify the membrane-disrupting activity of individual Aβ aggregates, we developed a single-vesicle assay that reports aggregate-induced damage with single-aggregate resolution. Nanoscale vesicles composed of neuronal membrane-mimicking lipids were loaded with a dextran-conjugated Ca2+-sensitive fluorogenic dye (Cal-520) within their lumens. The vesicles were then immobilised in a Ca2+-containing buffer on a coverslip (Fig. 1a). In this configuration, any disruption of the vesicle membrane by Aβ aggregates allows external Ca2+ to enter the lumen, where it binds to the encapsulated Cal-520 and generates a fluorescence signal that directly reports aggregate-induced damage. We quantified this increase in fluorescence from hundreds of individual vesicles simultaneously using total internal reflection fluorescence (TIRF) microscopy. Because each vesicle functions as an independent reaction chamber with an attoliter-scale (10−18 L) volume, and concentrations were chosen such that at most one Aβ aggregate encountered a vesicle at any given time, this design allowed us to quantify the damaging potential of hundreds of aggregates simultaneously with single-aggregate resolution (Fig. 1b).

Because our aim was to establish a regime where we could count the exact number of Ca2+ ions entering after an Aβ aggregate ‘hits’ the vesicle with single-ion resolution and use this signature as a fingerprint to distinguish different Aβ species, we determined the number of dye molecules encapsulated within each vesicle. We measured vesicle diameter in situ using super-resolution PAINT (Point Accumulation for Imaging in Nanoscale Topography) with Nile Red (Fig. 1c). Stochastic binding of Nile Red to vesicle membranes produced ‘blinking’ events that we localised with sub-diffraction precision (Fig. 1d), allowing us to estimate vesicle diameters down to ~ 50 nm (Supplementary Fig. 1a). Then we estimated the total number of dye molecules present within each vesicle by multiplying the calculated volume by the known encapsulated Cal-520 dye concentration used during vesicle preparation. To determine the maximal fluorescence signal for each vesicle, we added ionomycin, a Ca2+ ionophore, at the end of the experiment to saturate the lumen with Ca2+ -ion (Fig. 1e). This ensured that all Cal-520 molecules were bound to Ca2+, generating the maximal fluorescence signal and providing a direct correlation between dye number and fluorescence intensity. Cal-520 fluorescence was linear across the working concentration range, confirming the absence of self-quenching artifacts (Supplementary Fig. 1b). The measured fluorescence scaled proportionally with vesicle size, as confirmed by calculating the encapsulation efficiency (EE), the ratio of measured fluorescence intensity to theoretical dye content, for vesicles of different diameters. Consistent with previous reports37, the EE was close to unity for vesicles with diameters below 100 nm but declined for larger vesicles (Fig. 1f). Therefore, vesicles between 50 and 100 nm in diameter were selected for subsequent analyses.

We next exposed vesicles to Aβ42 aggregates and monitored for 20 min using time-lapse imaging. We observed discrete, stepwise increases in fluorescence, indicating transient bilayer disruptions followed by bursts of Ca2+ -ion influx triggered by damaging Aβ42 (Fig. 1g–i). These findings suggest that Aβ42 intermittently compromises neuronal membrane-mimicking bilayer integrity and permits bursts of Ca2+ -ion entry, as shown in earlier work38,39. By correlating total fluorescence intensity with the number of encapsulated dye molecules, we calculated the number of Ca2+ ions entering a vesicle after each hit (Fig. 1j). Each discrete fluorescence step corresponded to a single permeation event that produced a burst of Ca2+ -ion entry with single-ion resolution. Moreover, the assay is sensitive enough to detect even an influx of a single Ca2+ ion (Supplementary Fig. 2), because the vesicle lumen is on the order of attoliters and the entry of a single Ca2+ ion produces a micromolar-scale concentration change that lies well above the dissociation constant (Kd) of our Ca2+ -ion reporter. When we repeated the measurements under low aggregate concentrations, where only ~ 10–20% of vesicles encounter aggregates, the Ca2+-ion influx distributions remained unchanged, supporting the interpretation that the discrete Ca2+ -ion influx signatures arise predominantly from single aggregate-vesicle interactions (Supplementary Fig. 3). The frequency of hits reflects the abundance, whilst the magnitude of influx per hit quantifies the damaging capacity of each species. Together, these parameters provide quantitative measures of the total damaging effect of a sample (Fig. 1k).

Identification of two classes of functionally distinct Aβ42 intermediates

To track the emergence and evolution of the membrane-disrupting species during aggregation, we applied our assay to recombinant Aβ42 monomers (3 μM) undergoing fibrillization. Aggregation was monitored using Thioflavin T (ThT) fluorescence (Fig. 2a). In parallel, identical aggregation assays were performed without ThT, and aliquots were sampled at defined time points to measure membrane-disrupting activity. In the lag phase of aggregation (e.g., at 15 and 50 min), prior to any fibril formation detected by ThT, the Ca2+ -ion influx distributions were single-peaked at ~ 5 ions per event, defining a population we term class A intermediates (Fig. 2b). As aggregation continued, the frequency of class A events increased and peaked at the midpoint of aggregation (85 min), and declined thereafter, indicating that these class A species are transient and likely consumed or transformed during fibril elongation. By 480 minutes, the influx distribution again became single-peaked, centred at 2-3 Ca2+ ions per event (Fig. 2b). These relatively less-damaging species, designated as class B, first emerged at the end of the lag phase. Their abundance peaked at ~ 150 min, and gradually decreased thereafter. Despite this decline, class B aggregates remained detectable throughout the remainder of the aggregation process. In addition to these dominant A and B species, we observed a distinct influx distribution at the very early stages of aggregation (3 minutes), characterised by its low-magnitude signals with a peak at one ion per event and a fast decay. We refer to this rare, early-forming, and weakly damaging distribution as class C. Each class is defined by its characteristic distribution of Ca2+ -ion influx events, and the peak positions (5, ~ 2-3, and 1 ions per event for Class A, Class B and Class C, respectively) are used only as convenient descriptors throughout the manuscript.

These class C species were present at low levels and did not significantly contribute to membrane activity at later time points. To confirm the robustness of these classifications, we repeated Aβ42 aggregation at different concentrations (2 and 5 μM) and observed similar patterns of Ca2+-ion influx peaks (Supplementary Fig. 4a–d) corresponding to class A and class B intermediates. We also plotted the number of events at each Ca2+ -ion influx value over time normalised to 1 (Supplementary Fig. 5). The resulting curves for low and high Ca2+ -ion influx values clearly revealed two distinct kinetic behaviours corresponding to the rise and decline of our proposed class A and B species over the course of aggregation, further justifying this classification approach.

To probe the underlying pathways that lead to the formation of these toxic species, we aggregated Aβ42 in the presence of varying concentrations of BRICHOS R221E monomer, a chaperone known to slow down aggregation by specifically inhibiting secondary nucleation40,41 (Fig. 2c). For each condition, samples were collected at the half-time (t1/2) of the corresponding aggregation trace. We used t1/2 because it provides a reproducible reference point at which fibril mass reaches ~ 50% of its final level, enabling comparison of aggregate populations at an equivalent mechanistic stage of fibril formation despite differences in aggregation rate. The addition of BRICHOS at the start of the aggregation led to dose-dependent suppression of both class A (peaked at5 Ca2+ ions) and class B (peaked at ~ 2-3 Ca2+ ions) distribution (Fig. 2d). At 10% molar equivalents relative to monomeric Aβ42, BRICHOS reduced the class A peak by ~ 75%, and at 100% BRICHOS, it was completely eliminated. In contrast, class B events remained detectable even at the highest BRICHOS level (100%) (Fig. 2d). Because this assay detects only membrane-disruptive intermediates and not nondamaging ones, the reduction in events reflects a selective loss of class A and class B species or a loss of their damaging capacity in the presence of BRICHOS, rather than implying that intermediates dissociate or that aggregation proceeds without intermediate formation. By comparison, class C events were unaffected, even at 100% BRICHOS concentration (Fig. 2e). This behaviour cannot be attributed to a direct effect of BRICHOS on vesicle integrity, as this chaperone alone does not induce Ca2+ -ion influx in the absence of Aβ (Supplementary Fig. 6). Although the membrane-mimicking vesicles used here do not fully capture native membrane complexity, their controlled composition enables quantitative single-aggregate comparison of membrane-disrupting activity. Future studies using native membranes will be needed to determine how physiological membrane composition modulates the function of each intermediate class.

We then asked whether this BRICHOS R221E-mediated suppression of membrane-disrupting class A and B intermediates translates into neuroprotection. To test this, human pluripotent stem cell (hPSC)-derived cortical neurons (Fig. 2f) were exposed to Aβ42 aggregates collected at the midpoint of aggregation, when both class A and class B species are present. Aggregates were prepared either in the absence or presence of BRICHOS R221E, and neuronal connectivity was measured by quantifying neurite length every 6 h over five days using live-neuron imaging (Fig. 2g). This neurite connectivity assay parallels disruption of long-term potentiation, a well-established cellular correlate of learning and memory42. Aβ42 intermediates caused progressive neurite shortening, whereas BRICHOS R221E treatment conferred dose-dependent protection (Fig. 2h). Together, these results establish that Aβ42 aggregation generates distinct classes of neurotoxic intermediates with characteristic damaging profiles, and that their targeted suppression provides direct protection to human neurons.

Distinct Aβ42 intermediates have distinct kinetic signatures and damaging potential

At each stage of Aβ42 aggregation, the Ca2+ -ion influx distributions showed peaks at ~ 2–3 ions, ~ 5 ions, or both, corresponding to class B and class A species with distinct kinetic behaviours. To quantify these species accurately, we first defined pure fingerprints for each class. For class A, we selected the end of the lag phase (50 min for 3 μM Aβ42), when the influx distribution showed only the ~ 5 ion peak. This profile also closely resembled that at 15 min, early in the lag phase but without the small contributions from the rare class C species (Fig. 3a). For class B, we used the aggregation half-time in the presence of 100% BRICHOS R221E, a condition that completely suppressed class A activity and thus yielded a pure class B profile. This distribution also corresponds to the end of aggregation, when only class B species remained (Fig. 3b).

We then fitted the Ca2+ -ion influx distributions at each aggregation time point to a linear superposition of these fingerprint distributions. We found that, at every stage, membrane activity could be explained by either one species alone or by a mixture of the two, and the overall distributions can be described as linear combinations of their contributions (Fig. 3c), accounting for the full spectrum of Aβ42-induced membrane-disrupting activity. From this fitting, we quantified the contribution of class A and class B events at different aggregation time points, estimated their relative abundance and tracked their evolution across the aggregation trajectory (Fig. 3d). Class B aggregates followed a classical oligomerisation pattern: their abundance increased after the lag phase, peaked during the growth phase and declined once fibril formation saturated. In contrast, class A aggregates appeared earlier, peaking before substantial fibril formation. This timing is inconsistent with a pathway that follows fibrillization and instead indicates that class A species originate through a distinct kinetic route.

Finally, we fitted the Ca2+-ion influx distributions obtained in the presence of BRICHOS using the same fingerprint superposition to examine how it modulated the relative abundance of the two classes (Fig. 3e). The fits showed strong differential sensitivity: class A was completely suppressed by BRICHOS, whereas class B remained detectable even at the highest concentration tested (Fig. 3f). This differential response further supports that class A and class B arise through distinct aggregation pathways.

To further probe whether these functional classes correspond to physically distinct assemblies, we performed super-resolution imaging of aggregates collected at 50 and 180 min, which revealed that Class A aggregates are smaller on average than Class B (Supplementary Fig. 7a–d). We also analysed samples collected at 85 minutes, when both classes are present, using sucrose gradient fractionation. In these experiments, lower-density fractions were enriched in class A species, whereas higher-density fractions were enriched in class B aggregates (Supplementary Fig. 7e–g). ANS fluorescence analysis further showed a blue-shifted emission spectrum in the Class A-enriched fraction, suggesting greater exposed hydrophobicity (Supplementary Fig. 7h). Together, these data suggest that Class A and Class B are not only functionally distinct, but also differ in size and surface hydrophobicity.

Global kinetic analysis reveals distinct origins of intermediates

We next sought to contextualise the emergence of class A and class B species within the broader kinetics of Aβ42 aggregation. To this end, we performed global kinetic fitting of Aβ42 fibril formation data alongside the extracted kinetic curves for both species across multiple monomer concentrations (Fig. 4a). We first applied a state-of-the-art model in which all intermediates are predominantly formed via secondary nucleation (Fig. 4b), consistent with canonical Aβ42 oligomerisation mechanisms and previously reported rate constants43,44. This model failed to capture the early rise in class A species across concentrations (Fig. 4c) but adequately described the kinetics of class B species (Fig. 4d). We therefore tested an alternative model in which a much larger proportion of class A species are formed via primary rather than secondary nucleation, in contrast to what is observed for fibrils43 or directly measured oligomers44. This model produced substantially better agreement with the experimental data, particularly at early time points when class A species are most abundant.

Together, these analyses indicate that class A and class B species arise through distinct kinetic routes and have important implications for their connection to fibrils. Since fibrils are formed by primary and secondary nucleation in the same relative proportions as class B species, this implies that they are likely on-pathway intermediates of fibril nucleation and can be identified with the oligomeric intermediate species detected previously44. On-pathway in this context means that a substantial fraction of the overall nucleation flux passes through these intermediates45. It does not mean that most class B oligomers ultimately become fibrils; like all amyloid oligomers, both on- and off-pathway species are typically transient and predominantly dissociate back into monomers45,46. By contrast, class A species can contribute only minimally to fibril formation; if they did, a much larger share of fibrils would arise via primary nucleation. Consistent with this, we found that BRICHOS strongly inhibits class A species, and yet has no detectable impact on fibril primary nucleation47. This implies that class A species are off-pathway, contributing little to fibril formation45 (Fig. 4e).

When both populations were combined and treated as a single intermediate pool, as in earlier studies44, the model underestimated aggregate formation during the lag phase (Fig. 4f). Although this by itself does not prove the existence of two discrete intermediates, it exposes a fundamental inconsistency with a purely on-pathway mechanism.

Alzheimer’s brain contains the same two classes of intermediates

To assess the disease relevance of the intermediates of Aβ42 aggregation identified in vitro, we next asked whether class A and class B species are also present in the human brain. We examined postmortem (PM) frontal cortex tissue from three sporadic AD patients. In each case, we confirmed plaque pathology by immunohistochemistry using an Aβ40-specific antibody (EPR23712-2, Abcam) and an Aβ42-specific antibody (21F12, Absolute Antibody) (Fig. 5a and Supplementary Fig. 8). We found that more than 90% of amyloid plaques were Aβ42-immunoreactive. Then, to isolate diffusible intermediates, we applied a gentle extraction protocol previously shown to release diffusible, neurotoxic Aβ species48,49. Both small Aβ fibrils25 and neurotoxic oligomers42,48 have been detected in this fraction. However, mature fibrils have been shown to contribute minimally to membrane damage50. Therefore, any membrane-disruptive activity detected in our assay is expected to reflect diffusible Aβ intermediates.

To determine the specific contribution of Aβ42, brain extracts were immunoprecipitated with the Aβ42-specific 21F12 antibody or an isotype control. Fractions depleted with the control antibody showed higher levels of Aβ42 aggregates than those treated with 21F12, as confirmed by single-aggregate imaging (Supplementary Fig. 9). Functionally, these control-antibody depleted fractions also induced pronounced neurite degeneration in hPSC cortical neurons (Fig. 5b) and showed strong membrane-disrupting activity (Fig. 5c). In contrast, fractions immunodepleted with 21F12 displayed lower toxicity in both readouts.

We next quantified the net contribution of Aβ42 by calculating the difference of distribution  between isotype control and 21F12-immunodepleted supernatanat fractions (Fig. 5d). The resulting distributions revealed peaks at ~ 2-3 and ~ 5 Ca2+ -ion influx events, matching the class A and class B species identified in the recombinant systems. Fitting these data to their established fingerprint distributions confirmed the presence of both class A and B intermediates in all three AD patients (Fig. 5e), although their relative abundances varied across cases (Fig. 5f). These findings demonstrate that the two neurotoxic classes defined in vitro are also found in AD brain, indicating the disease relevance of the intermediates captured by recombinant Aβ peptide.

Although these species are transient in in vitro aggregation reactions, their detection in PM brain tissue is not unexpected. Flash-frozen PM tissue captures a snapshot of the in vivo aggregation landscape, where continuous Aβ monomer production and ongoing assembly maintain a dynamic reservoir of transient intermediates. By contrast, in vitro reactions lack sustained monomer supply; once monomers are consumed, intermediates diminish as aggregation proceeds to completion.

In vivo suppression of intermediates correlates with therapeutic benefit

To explore whether suppression of Aβ42 intermediates in vivo contributes to therapeutic benefit, we turned to a well-characterised mouse model of AD. The APPNL-G-F knock-in mouse model expresses humanised Aβ carrying the Arctic (E22G) mutation and shows progressive amyloid pathology, neuroinflammatory response of glial cells and behavioural abnormalities22,51. Previous studies have shown that intravenous administration of BRICHOS R221E monomer in this mouse model reduces plaque burden, attenuates astrogliosis, and improves cognition52,53. However, the molecular basis of these benefits remains unclear. Using tissue from the same study, we confirmed reductions in plaque load (Fig. 6a) and astrogliosis (Fig. 6b) and asked whether these therapeutic benefits are linked to a suppression of Aβ intermediates identified in our in vitro experiments.

As a first step, we aggregated recombinant Arctic Aβ42 (Fig. 6c) and characterised intermediates formed in the aggregation pathway. Although Arctic intermediates caused stronger membrane disruption than wild-type, consistent with the reported higher membrane-disrupting potency of Arctic Aβ42 attributed to increased transmembrane stability54, the overall pattern remained similar. At the midpoint of aggregation, we observed two peaks of Ca2+ -ion influx similar to class A and class B species. The addition of BRICHOS R221E suppressed both classes of intermediates, with a stronger effect on class A species (Fig. 6d), mirroring its impact on wild-type Aβ42. Thus, despite causing more membrane damage, Arctic Aβ42 displays the same pattern of Ca2+ -ion influx and similar sensitivity to BRICHOS.

We next asked whether these intermediates contribute to the attenuated pathology observed in APPNL-G-F mice when treated with BRICHOS R221E52,53. We collected Arctic Aβ42 aggregates from the aggregation midpoint (t1/2) and applied them separately to hPSC neurons or to iAstrocytes. In neurons, Arctic Aβ42 aggregates caused pronounced neurite shortening, which was substantially reduced when aggregates were prepared in the presence of BRICHOS R221E (Fig. 6e, f), consistent with the trend observed for wild-type Aβ42. A similar pattern was observed in iAstrocytes: Arctic Aβ42 induced robust secretion of the proinflammatory cytokines TNF-α, IL-1β, and IL-6, whereas aggregates prepared in the presence of BRICHOS elicited markedly lower cytokine release (Fig. 6g, h). Because astrocytic uptake is a major route of Aβ aggregate clearance55, we next examined whether BRICHOS affects this process. Surprisingly, uptake of Arctic Aβ42, quantified using an Aβ-specific 6E10 antibody, was comparable regardless of whether aggregates were formed with or without BRICHOS (Fig. 6i). However, when iAstrocytes were pre-treated with TNF-α for 24 h to mimic chronic inflammatory activation, uptake was significantly reduced (Fig. 6j). These results indicate that Aβ42 intermediates drive cytokine secretion and create a sustained pro-inflammatory environment that, in turn, impairs subsequent Aβ uptake. Thus, our results suggest BRICHOS reduces plaque pathology not by directly enhancing astrocytic uptake or producing Aβ42 aggregates that are inherently easier to internalise, but by preventing the formation of toxic intermediates that trigger pro-inflammatory signalling and ultimately hinder clearance.

To define how neurotoxic Arctic Aβ42 intermediates evolve, we first measured how Ca2+ -ion influx distributions change across the aggregation time course in a recombinant system (Fig. 7a). Two distinct populations again emerged: an early, high-influx distribution a peak at ~ 15 Ca2+ ions (Class A) and a later, lower-influx distribution with a peak at ~ 9 Ca2+ ions that dominated as fibrils began to form (Class B). As with wild-type, we extracted Arctic Aβ42-specific Ca2+ -ion influx fingerprint distributions for class A and B species using the influx profile from the lag phase and from the fibril half-time in the presence of BRICHOS R221E, respectively (Fig. 7b). These two fingerprints were sufficient to reconstruct the Ca2+ -ion influx distributions at all aggregation time points as linear combinations of Class A and Class B contributions.

We next asked whether these intermediates are suppressed in BRICHOS R221E–treated mice. Homogenates were prepared from the cortex and hippocampus of 13-month-old APPNL-G-F mice treated intravenously with either BRICHOS R221E or PBS (n = 6 per group)53. Single-aggregate imaging revealed a significant reduction in diffusible aggregates in BRICHOS-treated mice compared to PBS controls (Supplementary Fig. 10a–f). In PBS-treated animals, both brain regions showed robust Ca2+ -ion influx activity, which was significantly reduced after immunoprecipitation with the Aβ42-specific antibody 21F12, confirming that the activity was mediated by Arctic Aβ42 (Fig. 7c). Consistent with this reduction in Aβ42 aggregate abundance after BRICHOS treatment, membrane permeabilization assays revealed a marked decrease in membrane-disrupting activity (Fig. 7d, e and Supplementary Fig. 10g). Quantification showed that BRICHOS treatment markedly lowered the number of Ca2+ influx events in both hippocampal homogenates (decreasing from 9942 in PBS-treated samples to 6528 in BRICHOS-treated samples) and cortical homogenates (decreasing from 14810 to 8357) (Fig. 7d, e). Similar to our observations from the recombinant system, two distinct Ca2+ -ion influx peaks corresponding to Class A and Class B intermediates were detected in brain homogenates from both PBS- and BRICHOS R221E-treated mice across hippocampus (Fig. 7d) and cortex (Fig. 7e). Fitting these brain-derived distributions with the Class A and Class B basis functions showed that both species were present in PBS-treated mice, whereas BRICHOS R221E treatment reduced the abundance of both classes (Fig. 7f). These results further demonstrate that both classes of Aβ42 intermediates are present in vivo and indicate a molecular link between suppression of both intermediate species and the observed therapeutic effects.

Discussion

In this work, we demonstrate two mechanistically distinct Aβ42 aggregation routes that generate transient, neurotoxic intermediates with different damaging potential. Using a digital single-molecule assay, we defined these two classes of species across recombinant systems, an AD knock-in mouse model, and human Alzheimer’s brain. One class arises early during the lag phase, displays strong membrane-disrupting potency, and appears off-pathway relative to fibril formation. The other emerges later, more closely aligned with fibril precursors, and exhibits lower acute toxicity, but accumulates in greater numbers and thereby contributes substantially to the overall toxic burden. Suppression of both classes in vivo reduced plaque deposition, neuronal injury, and gliosis. These findings establish the physiological relevance of distinct aggregation routes and provide a mechanistic framework for understanding how early-stage Aβ intermediates drive AD progression.

In vitro studies of recombinant Aβ have revealed a wide spectrum of aggregates that differ in size (dimers, trimers, hexamers), structure (parallel/antiparallel β-sheets, β-hairpin assemblies), and morphology (globular and annular oligomers) formed en route to fibril formation56,57. Complementary evidence from human studies also indicates that Aβ intermediates with diverse morphologies exist in vivo and act as potent drivers of neurodegeneration24,58,59. Yet these species remain difficult to characterise because they are rare, transient, and heterogeneous60. As a result, several key questions remain unresolved: Are the intermediates observed in vitro the same species that are present in the human brain? Do they follow similar formation pathways in vivo? Is their mode of formation directly coupled to specific pathological functions? If so, can these insights be harnessed to design therapies that selectively neutralise pathogenic intermediates at the early stages of disease, when intervention is likely to be most effective?

Our work addresses these questions by introducing a single-molecule assay that integrates super-resolution imaging with functional readouts to classify Aβ assemblies at the single-species level by their ability to disrupt membranes and drive calcium dyshomeostasis. Since we observe Ca2+ -ion influx as transient events, this suggests that Aβ intermediates induce short-lived membrane permeabilisation, potentially through temporary pore formation or local membrane damage. These disruption processes represent key pathways through which Aβ contributes to synaptic failure and neuronal loss in Alzheimer’s pathology31–36. Our approach enables detection of rare and transient intermediates, disentangles coexisting species with distinct functions that remain hidden in ensemble measurements, and, when integrated with kinetic modelling, directly links formation pathways to functional outcomes. Such precision is essential because rare but highly toxic assemblies may arise through specific aggregation routes and exert a disproportionate impact on disease progression. Notably, this assay is not restricted to Aβ or to a specific lipid composition, and its modular design makes it broadly adaptable to other misfolding proteins involved in a wide range of protein-aggregation disorders.

Applying our framework to recombinant Aβ42 aggregation revealed a bifurcated kinetic trajectory defined by two mechanistically and functionally distinct intermediates (Fig. 8). Class A species emerge transiently during the lag phase, coinciding with primary nucleation or early oligomeric association steps, and display potent membrane-disrupting activity. Their transient nature reflects intrinsic molecular properties rather than abundance, and their short-lived nature likely explains why they have been largely missed in bulk measurements of soluble oligomers44. Class B intermediates, by contrast, accumulate during fibril growth and decline as fibril formation plateaus. Their temporal profile and sensitivity to fibril-surface inhibition are consistent with species generated through secondary nucleation, in line with classical models of fibril-catalysed oligomer formation and direct oligomer measurements44. Furthermore, inhibition of class B species by BRICHOS R221E closely mirrored its suppression of fibril secondary nucleation, which is achieved by binding to and blocking catalytic sites on fibril surfaces61,62. Although BRICHOS R221E changed the timing and magnitude of Aβ42 aggregation, the temporal profile of membrane-damaging species was preserved across conditions, with activity rising during early aggregation, peaking around the aggregation midpoint, and declining during the plateau phase (Supplementary Fig. 11). Supporting this, when we bypassed primary nucleation by seeding Aβ42 aggregation with pre-formed fibrils63, the lag phase was largely abolished and the reaction generated predominantly Class B species, with minimal Class A detection (Supplementary Fig. 12).

In contrast, Class A intermediates were reduced more strongly than expected from secondary nucleation suppression alone. These species begin to appear during the lag phase, when primary nucleation dominates, suggesting that their formation is not predominantly driven by fibril-surface-catalysed secondary nucleation. Consistent with this, the small molecule M1 previously shown to suppress Aβ42 secondary nucleation, selectively reduced Class B species, whereas Class A species remained the dominant membrane-damaging population (Supplementary Fig. 13). Moreover, when pre-formed aggregates collected at a time point containing both Class A and Class B species were incubated with BRICHOS R221E, the membrane-disrupting activity of Class A, but not Class B, was strongly reduced (Supplementary Fig. 14). These findings suggest that BRICHOS attenuates Class A activity after formation, likely through direct binding to these off-pathway intermediates, as reported previously64,65, rather than by preventing their initial generation. Together, these data support a model in which Class A and Class B species arise through distinct pathways: Class A represents an off-pathway, soluble membrane-active intermediate, whereas Class B is more closely linked to fibril-dependent secondary nucleation.

Distinct Aβ aggregates with different membrane-disrupting activity have been described previously and attributed to differences in hydrophobicity and epitope exposure20,24. Previous studies have also identified intermediate populations, some able to convert into fibrils and others that remain fibril-incompetent66,67, but the mechanistic basis of this distinction has been unclear. Our work provides the missing framework by positioning class A and class B within defined aggregation pathways and showing that their kinetic origins dictate both their toxic potency and whether they ultimately progress into fibrils. Together, these findings establish a mechanistic distinction: class B species follow the canonical fibril-surface generated oligomer route, whereas class A aggregates arise through an independent nucleation pathway. However, we cannot exclude that a small fraction of Class A species may enter fibril-forming pathways. We therefore interpret Class A as predominantly off-pathway and Class B as more closely linked to fibril-dependent secondary nucleation. Such on- and off-pathway intermediates have been reported for Aβ and other amyloidogenic proteins68, and species from both routes can contribute to toxicity and have been detected within Aβ plaques69. By contrast, class C species contribute minimally to overall membrane-disrupting activity only at the earliest stages of aggregation, and their formation route and molecular identity remain unclear.

After defining these aggregation routes in vitro, we next examined whether they also operate in the human brain to give rise to intermediates that contribute to neurodegeneration. Analysis of three AD brain extracts revealed diffusible, membrane-disruptive Aβ42 species corresponding to both class A and class B intermediates in every case examined, although their relative distribution varied between individuals. This variability may potentially reflect differences in disease stage, progression rate or genetic background. The coexistence of both intermediates across all cases supports a model in which primary and secondary nucleation pathways converge to drive pathogenic assembly in the human brain. Functionally, these species impaired human cortical neuronal connectivity, and selective immunodepletion of Aβ42 significantly mitigated this deficit. Because this neurite outgrowth assay parallels deficits in hippocampal long-term potentiation42, a well-established correlate of learning and memory, these results implicate these Aβ42 intermediates as direct contributors to synaptic failure and cognitive decline in AD. Although the cohort size is limited, these data provide initial evidence that both intermediates are present in human AD brain, while larger studies will be required to define how their relative abundance varies across disease stages and patient populations. Together, these results establish a mechanistic bridge between aggregation routes defined in model systems and intermediates that persist in the human Alzheimer’s brain and highlight these intermediates as key drivers of neurodegeneration and promising targets for therapeutic intervention.

We put this therapeutic idea to the test in the APPNL-G-F mice, which recapitulate key features of AD22,51. In BRICHOS R221E monomer-treated mice, plaque load was reduced, astrogliosis was attenuated, and cognitive function improved52,53,70, but the mechanistic basis for these improvements has remained unclear. BRICHOS does not block fibril formation outright; instead, it alters aggregation pathways, and how this rerouting translates into reduced pathology has not been resolved. To address this, we analysed mouse brain homogenates and found that untreated (PBS-treated) mice contained both class A and class B intermediates, whereas BRICHOS-treated mice showed a marked reduction in the abundance of both populations. Class A intermediates were suppressed more strongly than class B, consistent with the differential sensitivity observed in vitro. These data indicate that BRICHOS acts by reshaping the oligomeric landscape in vivo and lowering the abundance of toxic intermediates.

These findings further show how eliminating intermediates attenuates multiple pathological axes simultaneously. At the neuronal level, Aβ42 intermediates exert direct toxicity by impairing connectivity; their suppression improves neurite integrity. In plaque formation, at least one class of intermediates serves as a precursor to fibrils and fuels plaque growth; reducing these intermediates decreases the pool of fibril-forming species and lowers plaque burden in vivo. Intermediates also induce inflammatory responses. Arctic Aβ42 intermediates triggered secretion of proinflammatory cytokines, providing a molecular explanation for the astrocytic activation observed in APPNL-G-F mice. Exaggerated cytokine signalling causes synaptic and neuronal damage71 and sustains further reciprocal activation of microglia and astrocytes72. In addition, prolonged cytokine exposure impaired astrocytic uptake of Aβ, a major route of aggregate clearance55, paralleling earlier observations in microglia73,74 and explaining how intermediates can indirectly accelerate plaque accumulation. Finally, cytokine signalling also enhances neuronal Aβ production75,76, further amplifying aggregation burden and sustaining pathology. These data provide proof-of-principle that suppression of damaging Aβ intermediates in vivo correlates with reductions in disease-associated pathology.

Together, these processes converge into a self-reinforcing pathogenic cycle in which Aβ intermediates simultaneously drive neuronal injury, plaque accumulation, and pro-inflammatory signalling, each step further amplifying Aβ production and aggregation. Breaking this cycle requires suppressing intermediate formation at its point of origin, which reduces toxic assemblies, preserves neuronal health, limits inflammation, lowers plaque burden, and ultimately improves functional outcomes. These findings provide strong mechanistic evidence that early nucleation routes constitute a key actionable bottleneck through which the core pathology of AD can be directly intercepted.

Although these results advance our understanding of Aβ aggregation in vivo, several caveats remain. First, our assay detects species capable of membrane-disruption, so intermediates that do not perturb membranes may escape detection, and some species may continue to evolve under the assay conditions. This selectivity, however, ensures that we capture only membrane-damaging aggregates, which are strongly implicated in AD31–33. Second, although our data show that these intermediates are functionally and biophysically distinct, higher-resolution structural analyses will be required to define how their molecular architectures give rise to their distinct activities. Third, post-translational modifications of Aβ42, particularly N-terminal truncation and pyroglutamate formation (AβpE3), are found in the AD brain77. In our study, AβpE3 produced two classes of intermediates with membrane-permeabilising potency (Supplementary Fig. 15), similar to wild-type Aβ42. Because the C-terminus of Aβ42, a key determinant of membrane permeabilization20, remains unchanged in AβpE3, this likely accounts for their similar membrane-disrupting profiles to wild-type. The fact that AβpE3, like wild-type Aβ42, yielded two classes of species further supports the idea that nucleation-based aggregation routes operate in the human brain. Fourth, other Aβ isoforms are present in vivo and could complicate interpretation, most notably Aβ40, which is 10–20 times more abundant in healthy brain78. However, in regions most affected in AD, such as the cortex studied here, Aβ42 is predominant18,19, and we found Aβ42 levels were at least ten times higher than Aβ40 in both soluble aggregates (Supplementary Fig. 16) and insoluble plaques (Supplementary Fig. 8). Finally, while suppression of Aβ intermediates correlates with reduced plaque pathology and pro-inflammatory signalling in human cell models and Alzheimer’s mouse models, the contribution of these species to pathology extends beyond the mechanism examined here, and these findings should therefore be interpreted within the broader context of multiple Aβ-mediated pathogenic mechanisms79.

In summary, our findings demonstrate that toxic Aβ oligomeric intermediates are real, causal, and targetable in the Alzheimer’s brain. We define a mechanistic continuum between Aβ aggregation, plaque deposition, and neuronal toxicity across recombinant systems, mouse models, and the Alzheimer’s brain. By resolving aggregation pathways at the level of mechanism and function, our findings provide a framework for understanding how modulation of specific routes may limit the formation of toxic intermediates and contribute to reduced neuronal and glial pathology and plaque burden in model systems, supporting a testable strategy toward disease modification in AD. Detecting the same neurotoxic intermediates in Alzheimer’s brain as in our model systems supports the disease relevance of these models and highlights their value as platforms for therapeutic screening. More broadly, our framework establishes early molecular events as central drivers of disease progression and offers a generalisable strategy to intervene at the level of intermediate formation rather than downstream pathology, providing a path toward true disease-modifying therapies for early-stage AD and other protein-aggregation disorders.

Methods

Ethical statement

We complied with all relevant ethical regulations. All studies involving human and animal tissue were conducted in accordance with institutional guidelines and approved by the appropriate ethical review committees. Postmortem human brain tissue was obtained from the Sheffield Brain Tissue Bank, University of Sheffield, which operates under Research Tissue Bank approval from Wales Research Ethics Committee 5 (24/WA/0052). Informed consent was obtained from donors during life or from their next of kin. The astrocyte line was obtained through the MODEL-AD research study (Research Ethics Committee reference 16/YH/0155). All animal handling and procedures followed the comparative medicine guidelines of Karolinska Institutet, and the experiments were approved by the Stockholm Ethical Committee for Animal Experiments (15949-2020).

Recombinant Aβ42 expression and purification

Recombinant Aβ42 was expressed from the pET-Sac-Aβ(M1-42) plasmid (Addgene, 71875) transformed into E. coli BL21 Star™ (DE3)pLysS One Shot™ Chemically Competent cells (ThermoFisher, C602003), as previously published80. Briefly, single colonies were selected on LB-agar supplemented with 100 µg/mL ampicillin (ThermoFisher, J60977.06) and expanded overnight in 1 L LB broth containing 100 µg/mL ampicillin. Then protein expression was induced with 1 mM IPTG (Neo Biotech, NB-45-00030-25g) at an optical density of OD600 = 0.45-0.5. Cells were incubated for 5 h at 37 °C with shaking at 200 rpm before being pelleted by centrifugation at 6000 × g for 10 min. Pellets were stored at − 20 °C until further processing. For lysis, pellets were resuspended in buffer containing 10 mM Tris/HCl (pH 7.5) and 1 mM EDTA (Merck, 324503), supplemented with cOmplete™ EDTA-free protease inhibitor cocktail (Sigma, 11873580001) and DNase I (200 ng/mL). Cells were lysed on ice by sonication (4 x 30 s), followed by centrifugation at 18,000 × g for 8 min. This procedure was repeated twice. The resulting pellets were solubilized in 8 M urea (Merck, U5378), 10 mM Tris/HCl (pH 8.5), 1 mM EDTA, and clarified by centrifugation at 18,000 × g for 5 min. The supernatant was diluted 1:4 in 10 mM Tris/HCl (pH 8.5) with 1 mM EDTA. Samples were applied to a gravity column (Cytiva, 19-5101-01) packed with 10 mL DEAE Sepharose beads (Cytiva, 17-0709-01), pre-equilibrated with 2 M urea, 10 mM Tris/HCl (pH 8.5), 1 mM EDTA. The flow-through was collected and reapplied once, followed by three column washes with 10 mM Tris/HCl (pH 8.5) and 1 mM EDTA. Aβ42 was eluted with 10 mM Tris/HCl (pH 8.5), 1 mM EDTA, 500 mM NaCl. Solid GuHCl (ThermoFisher, 24110) was added directly to a final concentration of 6 M and samples were filtered through a 0.2 µm membrane and incubated overnight at room temperature.

Size-exclusion chromatography (SEC) was performed using a HiLoad® 16/600 Superdex® 30 pg column (Cytiva, 28-9893-31) on an ӒKTA pure™ HPLC system, with 20 mM ammonium acetate and 0.2 mM EDTA (pH 8.0) as the running buffer. Fractions corresponding to the Aβ42 monomer peak were pooled, quantified using a Qubit™ Protein Broad Range Assay Kit (ThermoFisher, A50668), aliquoted, snap-frozen in liquid nitrogen, and stored at − 80 °C.

Arctic Aβ42 (E22G) expression and purification

The Arctic Aβ42 (E22G) variant was generated from the plasmid pET-Sac-Aβ(M1-42) (Addgene, 71875) by site-directed mutagenesis using the QIAGEN QuikChange II kit according to the manufacturer’s protocol, with XL10-GOLD ultracompetent cells (Agilent, 200315) as hosts. Mutation-specific primers (forward 5′-ctggtgttcttcgctggagacgtgggttctaac-3′; reverse 5′-gttagaacccacgtctccagcgaagaacaccag-3′) were designed to introduce the E22G substitution. Mutated plasmids were purified using the QIAGEN Plasmid Miniprep kit, and DNA concentration and quality were verified by spectrophotometry. Arctic Aβ42 (E22G) was then expressed and purified similarly to wild-type Aβ42.

Thioflavin T based aggregation assay

Aggregation kinetics of wild-type and Arctic Aβ42 were monitored using a Thioflavin T (ThT) fluorescence assay. Monomeric Aβ42 was diluted in phosphate-buffered saline (PBS) to 200 µM, aliquoted (50 µL), flash-frozen in liquid nitrogen, and stored at − 80 °C. For aggregation assays, aliquots were thawed and further diluted in PBS to the desired concentration. Reactions were prepared in 96-well half-area plates (Corning, 3881) with 20 µM ThT (Sigma, T3516) and incubated at 37 °C without shaking. Plates were sealed, and fluorescence was monitored in real time using a Clariostar Plus plate reader (BMG Labtech).

For seeded aggregation assays, preformed Aβ42 fibril seeds were generated separately by incubating monomeric Aβ42 in PBS at 37 °C overnight. The resulting samples were transferred to low-binding tubes and sonicated for 2 min in a bath sonicator at room temperature. Fresh monomeric Aβ42 was then prepared as described above and supplemented with preformed fibril seeds at 1.5% or 2.5% relative to the monomer concentration. Seeded aggregation assays were performed in PBS containing 20 µM ThT at 37 °C under quiescent conditions.

N-terminally truncated and pyroglutamate-modified Aβ (AβpE3) was purchased (AnaSpec, AS-29907-01) and used without further purification. Aggregation assays were performed under identical conditions as for wild-type Aβ42. For BRICHOS R221E experiments, wild-type Aβ42 was aggregated in the presence of the chaperone at 10%, 50%, or 100% molar equivalents relative to monomer. For post-aggregation BRICHOS treatment experiments, pre-formed Aβ42 aggregates were collected at the aggregation midpoint and incubated with BRICHOS R221E at 100% molar equivalent relative to Aβ42 monomer before analysis using the single-vesicle Ca2+ -ion influx assay. For small-molecule inhibition experiments, 3 μM Aβ42 was aggregated in the presence of N-(2-{[1,1′-biphenyl]-4-yl}ethyl)-3,4-dihydro-1H-2,3-benzoxazine-3-carboxamide (M1; Enamine, Z8478213987) at 0, 5, or 20 μM as previously mentioned81. Reactions were monitored with ThT fluorescence as described earlier.

Dual-labelled Aβ42 aggregation

Monomeric Aβ42 conjugated to the FRET partner dyes HiLyte Fluor 555 (AnaSpec, AS-60480-01) and HiLyte Fluor 647 (AnaSpec, AS-64161) was obtained commercially. The peptides were first dissolved in 10 mM NaOH and subsequently purified by HPLC using a Bio-Sep 2000 column. Fractions corresponding to the peptide peak were collected, and concentrations were determined from the absorbance of the attached dye. Donor- and acceptor-labelled monomers were then combined in a 1:1 ratio so that half of the molecules carried the donor dye and half the acceptor dye. Aggregation reactions were initiated at a final protein concentration of 3 µM and incubated at 37 °C under quiescent conditions in PBS in 96-well half-area plates. Aliquots were taken at the same time points as those used for aggregation reactions performed with unlabelled Aβ42.

Wide-field fluorescence imaging setup

Single-molecule and super-resolution imaging were performed using a custom-built microscope based on a Nikon Eclipse Ti2 body equipped with a Perfect Focus unit and Omicron Luxx lasers. Lasers were launched through a fibre coupler (KineFlex SM/PM Fiber), collimated with Zoom Fiber Collimators (Thorlabs C20FC-A), and passed through achromatic Quarter-Wave Plates (Thorlabs AQWP05M-600). Then the laser beam directed into the back focal plane of a 100x Plan ApoTIRF, 1.49 NA oil-immersion objective lens (Nikon). To achieve uniform illumination in epifluorescence and TIRF, a beam shaper (Asphericon, TSM25-10-LD-D-532) was integrated in the excitation path, resulting in less than 5% intensity variation across the imaging field73. Fluorescence emissions were collected through the same objective and separated using a dichroic beamsplitter. The light then passed through a specific set of optical filters for each fluorophore before being captured by an sCMOS camera (Photometrics Prime 95B). The whole setup was controlled by Micro-Manager 2.0.

Vesicle preparation and immobilisation on glass coverslips

Vesicles were prepared following established protocols with minor modifications82. Briefly, 35% PC (Avanti, 131101), 15% PE (Avanti, A85757), 8% PS (Avanti, A80534), 2% PI (Avanti, A85079) and 30% Cholesterol (Sigma, 700100 P) and 10% Ganglioside GM1 (Avanti, A86065) were dissolved in chloroform and methanol mixture (4:1 v/v) and then further mixed with biotinylated lipids (18:1-12:0 Biotin PC; Avanti, 860563) at a molar ratio of 100:1. The solvent was removed under vacuum in a desiccator overnight, and the resulting lipid film was rehydrated in HEPES buffer (pH 6.5) containing 100 µM Cal-520 (500 µM for Arctic Aβ). The suspension underwent at least five freeze-thaw cycles, alternating between dry ice and a water bath. To obtain vesicles of ~ 100 nm or smaller, the solution was extruded at least 10 times through a 100 nm pore-size polycarbonate membrane (Avanti). Unincorporated dye molecules were removed by size-exclusion chromatography using a Superdex™ 200 Increase 10/300 GL column on an ÄKTA Pure system (GE Life Sciences) with a flow rate of 0.5 mL/min.

For vesicle immobilisation, borosilicate glass coverslips (22 × 22 mm2, VWR 6310122) were sequentially cleaned by sonication in Milli-Q water and methanol, dried under nitrogen, and plasma-treated for 20 min in an argon plasma cleaner (Deiner Zepto). Frame-Seal incubation chambers (9 × 9 mm², Bio-Rad) were affixed to the coverslips, and surfaces were functionalised with a mixture of PLL-g-PEG and PLL-g-PEG-biotin (100:1, SuSoS AG; 1 g/L in HEPES buffer, pH 6.5) for 30 min. After washing with buffer, coverslips were incubated with NeutrAvidin (0.1 mg/mL in HEPES, ThermoFisher) for 15 min and washed again. Purified vesicles were then added to the chambers, incubated for 30 min, and excess vesicles were removed by three washes with reaction buffer.

Membrane permeabilisation assay

Membrane permeabilisation was measured using a single-vesicle Ca2+-ion influx assay adapted from established protocols50,83. Vesicle chambers were first exchanged into Ca2+-containing buffer, and baseline fluorescence images (Fblank) were acquired as 10 consecutive frames with 50 ms exposure per frame. These measurements provided the background intensity for each vesicle. Vesicles were then exposed to 1 nM Aβ42 (monomer equivalents), a concentration chosen to enable detection of discrete fluorescence increases corresponding to individual Ca2+-ion influx events within single vesicles within this time frame. After sample addition (Fsample), imaging continued once every minute for 20 min using the same acquisition scheme: one imaging burst per minute consisting of 10 frames with 50 ms exposure each. Between imaging bursts, the sample remained unilluminated to minimise photobleaching. Following this period, ionomycin (1 mg/mL; Cambridge Bioscience, 1565) was added to fully saturate vesicles with Ca2+, and imaging proceeded for a further 5 min using the same settings (Fionomycin). Relative total Ca2+-ion influx was calculated for each vesicle for each hit as:\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{{\rm{Ca}}}}^{2+}-{{\rm{ion}}}\,{{\rm{influx}}}=({{{\rm{F}}}}_{{{\rm{sample}}}}-{{{\rm{F}}}}_{{{\rm{blank}}}})\,/\,({{{\rm{F}}}}_{{{\rm{ionomycin}}}}-{{{\rm{F}}}}_{{{\rm{blank}}}})$$\end{document}Ca2+−ioninflux=(Fsample−Fblank)/(Fionomycin−Fblank)

For each vesicle, the fluorescence increment associated with an individual hit was calculated as ΔF_event = F_after − F_before, where F_before and F_after are the fluorescence levels immediately before and after the step. For analysis, the 10 frames within each imaging burst were averaged to obtain a single intensity value per burst. Under these conditions, most vesicles encounter Aβ aggregates and show discrete Ca2+ influx events without signal saturation. This averaging step improved the signal-to-noise ratio by reducing frame-to-frame fluctuations, allowing us to detect and quantify very small fluorescence changes corresponding to individual Ca2+-influx events. Cal-520 was excited at 488 nm, and emission was collected through a combination of a long-pass filter (BLP01-488R-25, Laser2000) and a band-pass filter (FF01-520/44-25, Laser2000) and imaged onto a Prime 95B sCMOS camera (Photometrics). Images were acquired at ~ 1–10 W/cm² power density with a scan speed of 20 Hz and 16-bit depth, using flat-field illumination with less than 5% intensity variation between the periphery and the centre of the image. Fields of view were selected automatically using a motorised stage controlled by Micro-Manager scripting to ensure unbiased data collection.

As controls, we incubated monomeric Aβ42 (1 nM) with lipid vesicles in solution for 30 min and did not observe detectable aggregate formation (Supplementary Fig. 17). Similarly, when pre-formed aggregates (1 nM) were incubated with lipid vesicles, the aggregate number remained unchanged. In addition, incubation of pre-formed Aβ42 aggregates (1 nM) in Ca2+ buffer alone showed no detectable change in aggregate number (Supplementary Fig. 18), indicating that neither the lipid membrane of the vesicles nor the Ca2+ buffer induces any artefacts under the experimental conditions.

Vesicle size estimation using point accumulation for imaging in nanoscale topography (PAINT)

To determine the maximum fluorescence capacity of individual vesicles, we first estimated vesicle size in order to calculate the number of dye molecules encapsulated per vesicle, given that the dye concentration during vesicle preparation was known. Although a cutoff filter was used during preparation, it did not guarantee uniform size, as smaller vesicles could still form and introduce variation in dye content. For instance, a 100 nm cutoff filter can yield vesicles of both 100 nm and 50 nm in diameter; because volume scales with the cube of diameter, a 100 nm vesicle contains approximately eight times more dye molecules than a 50 nm vesicle.

Therefore, vesicle diameters were determined in situ using point accumulation for imaging in nanoscale topography (PAINT) with transient binding of Nile Red to vesicle membranes. First, Nile Red was dissolved in dimethyl sulfoxide and then diluted to 1 nM in PBS immediately prior to imaging. Samples were imaged in through-objective TIRF mode using a 532 nm laser for excitation. Emission was collected through a dichroic mirror (Di02-R532-25 × 36). Image sequences of 1000 frames were acquired at 30 ms exposure. Localisations were extracted using the Peak Fit plugin for ImageJ/Fiji, and coordinates corresponding to individual vesicles were clustered using DBSCAN. Vesicle size was estimated by Gaussian fitting of the clustered localisations. Clusters containing fewer than 40 localisations were excluded as background. For calibration, fluorescence intensities obtained after ionomycin-induced Ca2+ -ion influx were correlated with vesicle size.

Analogue Ca2+-ion influx conversion to digital number of Ca2+-ion influx

To convert the analogue percentage readout of Ca2+-ion influx into absolute ion numbers, we calculated the number of dye molecules per vesicle from vesicle size data. The number of Cal-520 molecules per vesicle was estimated from the known encapsulated Cal-520 concentration and the vesicle volume, derived from PAINT-based diameter measurements. The theoretical dye number was then used to convert percentage influx values into ion counts according to:\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\rm{Number}}}\; {{\rm{of}}}\;{{\rm {Ca}}}^{2 + }-{{\rm{ion}}}\;{{\rm{entering}}}\; {{\rm{per}}}\;{{\rm{influx}}}\; {{\rm{event}}} \\={{{\rm{N}}}}_{{{\rm{dye}}}}{{\rm{x}}}({{{\rm{F}}}}_{{{\rm{after}}}}-{{{\rm{F}}}}_{{{\rm{before}}}})\,/\,({{{\rm{F}}}}_{{{\rm{ionomycin}}}}-{{{\rm{F}}}}_{{{\rm{blank}}}})$$\end{document}NumberofCa2+−ionenteringperinfluxevent=Ndyex(Fafter−Fbefore)/(Fionomycin−Fblank)where Ndye is the number of Cal-520 molecules per vesicle. The fractional values obtained were rounded to yield discrete Ca2+ ion counts. From these measurements, we calculated the number of dye molecules encapsulated in each vesicle. Experiments were conducted within a range where Cal-520 fluorescence intensity varied linearly with concentration, avoiding dye aggregation or quenching. We measured the encapsulation efficiency (EE), defined as the ratio between measured and calculated intensity, by plotting integrated Cal-520 fluorescence against vesicle volume. In our system, EE was close to unity for vesicles up to ~ 100 nm in diameter but decreased for larger vesicles. Therefore, vesicles were prepared with a 100 nm cutoff filter, and vesicles larger than this threshold were excluded from analysis. Effective vesicle diameters were then obtained by subtracting 10 nm from the measured diameters to account for bilayer thickness84. Similar to the previous report, we also found empty vesicles that did not contain any encapsulated dye37, and these were excluded from the analysis.

For control experiments, BRICHOS R221E alone was used at the same concentrations and incubated under the same conditions as in the Aβ42 aggregation reactions. The samples were incubated in the same well plate for the same duration before being added to vesicles immobilised on the coverslip. Membrane permeabilisation measurements were then performed using the same assay conditions and imaging procedure as described for the Aβ42 samples.

Expression and purification of Bri2 BRICHOS R221E monomer

The Bri2 BRICHOS R221E variant was prepared following a previously described protocol with minor modifications40,64. Site-directed mutagenesis was performed on the wild-type His6-NT*-Bri2 BRICHOS (residues 113–231) plasmid to introduce the R221E substitution. The construct was expressed in E. coli Shuffle T7 cells cultured in LB medium supplemented with 15 µg/mL kanamycin at 30 °C. Protein expression was induced with 0.5 mM IPTG at an optical density of 0.8, and cultures were incubated overnight at 20 °C. Cells were harvested by centrifugation (5000 × g, 20 min, 4 °C), resuspended in 20 mM Tris-HCl buffer (pH 8.0), lysed by sonication, and clarified by centrifugation (24,000 × g, 30 min, 4 °C). The NT*-BRICHOS fusion protein was purified by Ni-NTA affinity chromatography. To remove the His6-NT* solubility tag, thrombin digestion was performed during overnight dialysis, leaving an additional glycine and serine at the N-terminus of BRICHOS. A second Ni-NTA purification step was used to separate the His6-NT* fragment, and the flow-through containing BRICHOS was collected. Samples were heat-treated at 60 °C for 10 min and monomeric protein was isolated by size-exclusion chromatography on a HiLoad Superdex 75 pg column. The purified BRICHOS R221E monomer was concentrated to the desired level and stored at − 70 °C until use.

Endotoxin measurement in recombinant protein

Endotoxin levels in recombinant Aβ42 and Bri2 BRICHOS R221E monomer preparations were quantified using a commercial endotoxin detection kit (Thermo Fisher Scientific, Cat. No. A39553) according to the manufacturer’s instructions. Endotoxin levels were 0.03 EU/mg for recombinant Aβ42 and 0.05 EU/mg for Bri2 BRICHOS R221E monomer. Furthermore, recombinant Aβ42 was passed through an endotoxin removal spin column (ThermoFisher, Cat. No. 88274). The protein concentration was remeasured and aggregation kinetics were monitored using ThT fluorescence, which remained unchanged (Supplementary Fig. 19). To further assess potential functional consequences of endotoxin contamination, we analysed astrocyte-mediated cytokine secretion using the resulting aggregates, as astrocytes are highly sensitive to even minimal levels of endotoxin. Following endotoxin removal, we observed no differences in astrocyte cytokine responses compared with untreated samples (Supplementary Fig. 19).

Basis function fitting

Basis functions representing the Ca2+ ion influx distribution of class A, B and C intermediates were selected as described in the Results. These were then normalised to unity and fitted using a simple Python script to the overall Ca2+ ion influx distributions at each time point and condition using a standard nonlinear least squares method.

Kinetic modelling of fibril formation

The kinetics of fibril formation by Aβ42 can be well-described by a model featuring three reaction steps. These are primary nucleation of new fibrils with rate constant \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{n}$$\end{document}kn and reaction order with respect to monomer \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{c}$$\end{document}nc, fibril surface-catalysed secondary nucleation with rate constant \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{2}$$\end{document}k2 and reaction order with respect to monomer \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{2}$$\end{document}n2, and elongation with rate constant \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{+}$$\end{document}k+. The integrated rate law corresponding to this model was determined in ref. 85. In this model, fibril mass concentration (which is the quantity tracked by ThT fluorescence) is denoted \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$M\left(t\right)$$\end{document}Mt, and total initial monomer concentration is denoted \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{{\mbox{tot}}}$$\end{document}mtot. Setting the saturation concentrations to infinity, which are not needed here due to the relatively small concentration range of 2–5 µM, gives:1a\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$M\left(t\right)={m}_{{\mbox{tot}}}-{m}_{{\mbox{tot}}}{\left(1+\frac{{{\rm{\epsilon }}}}{c}\left({e}^{{{\rm{\kappa }}}t}+{e}^{-{{\rm{\kappa }}}t}-2\right)\right)}^{-c}$$\end{document}Mt=mtot−mtot1+ϵceκt+e−κt−2−c1b\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\rm{\kappa }}}=\sqrt{2{k}_{+}{k}_{2}{m}_{{\mbox{tot}}}^{{n}_{2}+1}}$$\end{document}κ=2k+k2mtotn2+11c\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\rm{\epsilon }}}=\frac{{k}_{n}{m}_{{\mbox{tot}}}^{{n}_{c}}}{{2k}_{2}{m}_{{\mbox{tot}}}^{{n}_{2}+1}}$$\end{document}ϵ=knmtotnc2k2mtotn2+11d\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$c=\frac{3}{2{n}_{2}+1}.$$\end{document}c=32n2+1.

Fitting the data with this model gives combined rate constants \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{n}{k}_{+}=3.01\cdot {10}^{-7}\,{{{\rm{\mu }}}{{\rm{M}}}}^{-{{{\rm{n}}}}_{c}}\,{{{\rm{mins}}}}^{-2}$$\end{document}knk+=3.01⋅10−7μM−ncmins−2, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{2}{k}_{+}=2.93\cdot {10}^{-4}\,{{{\rm{\mu }}}{{\rm{M}}}}^{-{n}_{2}-1}\,{{{\rm{mins}}}}^{-2}$$\end{document}k2k+=2.93⋅10−4μM−n2−1mins−2, and reaction orders \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{c}=1.74,{n}_{2}=1.37$$\end{document}nc=1.74,n2=1.37. The \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{2}$$\end{document}n2 value < 2 reflects partial saturation; Aβ42 secondary nucleation is known to saturate as \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{{\mbox{tot}}}$$\end{document}mtot approaches 6 µM. Fits were performed using an offline version of AmyloFit86, which is largely Python-based.

Kinetic modelling of oligomers

The basis function fitting produced kinetic curves for class A and B intermediates at 3 distinct initial monomer concentrations. Oligomer kinetics can be described with the following integrated rate laws46, where \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$S\left(t\right)$$\end{document}St denotes oligomer concentration:2a\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$S\left(t\right)={S}_{1}\left(t\right)+{S}_{2}\left(t\right)$$\end{document}St=S1t+S2t2b\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{1}\left(t\right)=\frac{{{{\rm{\alpha }}}}_{1}}{{{{\rm{\rho }}}}_{e}\left(t\right)}\left(\frac{c}{c-\epsilon}\right)^{c\, n_{o1}}\left({}_{2}{F}_{1}\left[c{n}_{o1},\frac{{{{\rm{\rho }}}}_{e}\left(t\right)}{{{\rm{\kappa }}}},1+\frac{{{{\rm{\rho }}}}_{e}\left(t\right)}{{{\rm{\kappa }}}},\frac{\epsilon {e}^{{{\rm{\kappa }}}t}}{\epsilon -c}\right]{-}e^{-\rho_e(t)t}{}_{2}{F}_{1}\left[c{n}_{o1},\frac{{{{\rm{\rho }}}}_{e}\left(t\right)}{{{\rm{\kappa }}}},1+\frac{{{{\rm{\rho }}}}_{e}\left(t\right)}{{{\rm{\kappa }}}},\frac{\epsilon }{\epsilon -c}\right]\right)$$\end{document}S1t=α1ρetcc−ϵcno1F12cno1,ρetκ,1+ρetκ,ϵeκtϵ−c−e−ρe(t)tF12cno1,ρetκ,1+ρetκ,ϵϵ−cwhere 2F1[a,b;c;z] is the Gaussian hypergeometric function.2c\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\begin{array}{c}{S}_{2}(t)=\frac{{{{\alpha }}}_{2}m_{{\mbox{tot}}}}{{\rho}_{e}(t)}\left(\frac{c}{c-\epsilon}\right)^{c\, n_{o2}}\left({}_{2}{F}_{1}\left[c{n}_{o2},\frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} ,1 + \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , \frac{\epsilon {e}^{{{\kappa }}t}}{\epsilon -c}\right]{-}e^{-\rho_e(t)t}{}_{2}{F}_{1}\left[c{n}_{o2} , \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , 1 + \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , \frac{\epsilon }{\epsilon -c}\right]\right)\\ -\frac{{{{\alpha }}}_{2}m_{{\mbox{tot}}}}{{{{\rho }}}_{e}\left(t\right)}\left(\frac{c}{c-\epsilon}\right)^{c\, (n_{o2} + 1)}\left({}_{2}{F}_{1}\left[c\left({n}_{o2} + 1\right) , \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , 1 + \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}},\frac{\epsilon {e}^{{{\kappa }}t}}{\epsilon -c}\right]{-}e^{-\rho_e(t)t}{}_{2}{F}_{1}\left[c\left({n}_{o2} + 1\right) , \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , 1 + \frac{{{{\rho }}}_{e}\left(t\right)}{{{\kappa }}} , \frac{\epsilon }{\epsilon -c}\right]\right)\end{array}$$\end{document}S2(t)=α2mtotρe(t)cc−ϵcno2F12cno2,ρetκ,1+ρetκ,ϵeκtϵ−c−e−ρe(t)tF12cno2,ρetκ,1+ρetκ,ϵϵ−c−α2mtotρetcc−ϵc(no2+1)F12cno2+1,ρetκ,1+ρetκ,ϵeκtϵ−c−e−ρe(t)tF12cno2+1,ρetκ,1+ρetκ,ϵϵ−c2d\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{{\rm{\rho }}}}_{e}\left(t\right)={k}_{e1}+{k}_{e2}M\left(t\right),$$\end{document}ρet=ke1+ke2Mt,

With M(t) is given by Equation (1a). The parameters α1, α2 are given by:3a\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{{\rm{\alpha }}}}_{1}=_{{\mbox{tot}}}^{{n}_{o1}}$$\end{document}α1=totno13b\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{2}={k}_{o2}{m}_{{\mbox{tot}}}^{{n}_{o2}}.$$\end{document}α2=ko2mtotno2.

\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{o1}$$\end{document}ko1 and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{o2}$$\end{document}ko2 are the rate constants for oligomer formation by primary and secondary nucleation respectively, and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{o1}$$\end{document}no1 and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{o2}$$\end{document}no2 the corresponding monomer reaction orders. \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{e1}$$\end{document}ke1 and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{e2}$$\end{document}ke2 are the rate constants for their disappearance by free and fibril-catalyzed dissociation respectively. Since dissociation must be catalysed by fibril surfaces comparably to oligomer formation, the additional relation holds:4\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{e1}={k}_{e2}\frac{{{{\rm{\alpha }}}}_{1}}{{{{\rm{\alpha }}}}_{2}}.$$\end{document}ke1=ke2α1α2.

If the oligomers are on-pathway to fibril formation, the rate constants from the 2-state model used to fit the fibril kinetic curves can be used to constrain the rate constants in the oligomer kinetic model using the following formulae from ref. 46:5a\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{k}_{+}k}_{o2}=\frac{{\left(2{k}_{+}{k}_{2}\right)}^{3/2}}{2{k}_{{\mbox{conv}}}}$$\end{document}k+ko2=2k+k23/22kconv5b\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{k}_{+}k}_{o1}=\frac{3{{k}_{n}k}_{+}{k}_{o2}}{2{k}_{2}}$$\end{document}k+ko1=3knk+ko22k25c\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{o2}={\left(3/2\right)}^{{n}_{2}+1}-1-{n}_{{\mbox{conv}}}$$\end{document}no2=3/2n2+1−1−nconv5d\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{o1}={\left(3/2\right)}^{{n}_{c}}-{n}_{{\mbox{conv.}}}$$\end{document}no1=3/2nc−nconv.

Using these relations, the rate equations can then be fitted to the oligomer kinetic curves, fitting the free parameters \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{{\mbox{conv}}},{n}_{{\mbox{conv}}},{k}_{e2}$$\end{document}kconv,nconv,ke2. The remaining parameters \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{k}_{+}k}_{o1},{{k}_{+}k}_{o2},{n}_{o1},{n}_{o2}$$\end{document}k+ko1,k+ko2,no1,no2 are then calculated using Eqs. (5) after fitting.

If an oligomer population is instead off-pathway, equation (5) do not apply, and the free parameters for fitting are instead \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{k}_{+}k}_{o1},{{k}_{+}k}_{o2},{n}_{o1},{n}_{o2},{k}_{e2}$$\end{document}k+ko1,k+ko2,no1,no2,ke2.

Again, fitting was performed using an offline implementation of AmyloFit86. (Note, since the number of vesicle binding events per nM of oligomers is unknown, the fitted ko1, ko2 are valid only up to an unknown constant of proportionality.) This yielded misfits in the case that class A intermediates are considered on-pathway. Good fits were instead obtained when treating class A intermediates as off-pathway, with the following combined rate parameters: \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{e2}=0.072\,{{{\rm{\mu }}}{{\rm{M}}}}^{-1}\,{{\mbox{mins}}}^{-1}$$\end{document}ke2=0.072μM−1mins−1, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{o1}=2.19,{n}_{o2}=0.606$$\end{document}no1=2.19,no2=0.606, and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{+}{k}_{o1}=3.98\,{{{\rm{\mu }}}{{\rm{M}}}}^{-{n}_{o1}}\,{{{\rm{mins}}}}^{-2}$$\end{document}k+ko1=3.98μM−no1mins−2, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{+}{k}_{o2}=325\,{{{\rm{\mu }}}{{\rm{M}}}}^{-{n}_{o2}-1}\,{{{\rm{mins}}}}^{-2}$$\end{document}k+ko2=325μM−no2−1mins−2.

By contrast, good fits were determined when class B intermediates were considered on-pathway, with the following rate parameters: ke2 = 0.00101 \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{{\upmu }}{{\rm{M}}}}^{-1}$$\end{document}μM−1 min−1, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${k}_{{{\rm{conv}}}}=2.68\cdot {10}^{-7}\,{{{\rm{\mu }}}{{\rm{M}}}}^{-{n}_{{{\rm{conv}}}}}\,{{{\rm{mins}}}}^{-1}$$\end{document}kconv=2.68⋅10−7μM−nconvmins−1, \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${n}_{{\mbox{conv}}}=0.44$$\end{document}nconv=0.44.

Human postmortem tissue case selection

Postmortem brain tissue was obtained from three individuals with clinically diagnosed and pathologically confirmed AD (age range: 71–92 years; two female and one male). To assess both amyloid plaque pathology and the presence of soluble Aβ assemblies, we focused on cortical regions that consistently display substantial Aβ deposition in AD. From each donor, two complementary tissue preparations were obtained from the same anatomical areas: (i) formalin-fixed, paraffin-embedded (FFPE) samples for immunohistochemical analysis, and (ii) flash-frozen tissue for biochemical extraction of diffusible aggregates. FFPE tissue is required for plaque quantification, whereas flash-frozen tissue preserves the native conformations of soluble Aβ species, which are altered by formalin cross-linking. To allow these parallel analyses, each brain was divided sagittally. One hemibrain underwent formalin fixation before dissection and paraffin embedding, while the opposite hemibrain was sectioned coronally and immediately flash-frozen in liquid nitrogen-cooled teflon-coated copper plates.

Immunohistochemistry of postmortem AD tissue

Formalin-fixed paraffin-embedded (FFPE) frontal cortex sections (5 µm) from three AD cases were used for immunohistochemistry with an antibody specific for Aβ40 (EPR23712-2, Abcam), Aβ42 (21F12, Absolute Antibody, Ab02391-23.0) and DAPI. Sections were dewaxed in xylene, rehydrated through a graded ethanol series, and incubated in 3% H2O2 in methanol for 10 min to quench endogenous peroxidase activity. Antigen retrieval was performed in citrate buffer (pH 6.0) by microwave at full power for 10 min, followed by a 3 min incubation in 100% formic acid. To block nonspecific binding, sections were incubated with 2.5% normal horse serum (VECTASTAIN® Elite ABC-HRP Kit, Mouse IgG, Vector Laboratories, PK-6102) for 30 min at room temperature. Sections were then incubated overnight at 4 °C with primary antibody (1:500 and 1:1000 dilution, respectively), followed by Alexa Fluor 488 (Thermo Fisher, A-11008) and Alexa Fluor 647 (Thermo Fisher, A-21236) secondary antibody for 1 h at room temperature. Slides were scanned at 20x magnification using a Hamamatsu NanoZoomer XR slide scanner.

Extraction of diffusible protein aggregates from human postmortem tissue

Diffusible Aβ-containing aggregates were extracted from postmortem human frontal cortex tissue using a previously described protocol with minor modifications48,49. Briefly, tissue samples were cut into ~ 300 mg pieces with a razor blade and incubated with 1 mL artificial cerebrospinal fluid (aCSF; 120 mM NaCl, 2.5 mM KCl, 1.5 mM NaH2PO4, 26 mM NaHCO3, 1.3 mM MgCl2, pH 7.4) at 4 °C for 30 min under gentle agitation. Samples were centrifuged at 2000 × g for 10 min at 4 °C, and the upper ~ 80% of the supernatant was collected and centrifuged again at 14,000 × g for 2 h at 4 °C. The resulting supernatant ( ~ 80% of volume) was dialysed for 72 h at 4 °C against a 50-fold excess of aCSF using a 2 kDa molecular weight cutoff (MWCO) Slide-A-Lyzer (Thermo Fisher), with buffer exchanges every 12 h. Dialysed extracts were aliquoted, snap-frozen, and stored at − 80 °C until use, avoiding additional freeze-thaw cycles.

Immunohistochemistry of APPNL-G-F mouse tissue

Brain tissue from APPNL-G-F knock-in mice that had been treated previously52,53, with either PBS or recombinant human BRICHOS R221E was used for analysis. In that study, animals received intravenous injections of PBS (vehicle) or BRICHOS R221E (10 mg/kg) twice weekly for 12 weeks beginning at 9 months of age, followed by a four-week treatment-free period. At the study endpoint, brains were dissected and processed for both immunohistochemistry and biochemical assays. For histological analyses, the hippocampus and retrosplenial cortex from the right hemispheres were immersion-fixed in 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer at 4 °C overnight, cryoprotected in 30% sucrose for 24 h, and transferred to cryoprotectant solution at − 20 °C. Coronal sections (20 µm) were cut and used for immunostaining. Sections were washed in PBS, permeabilised with 0.3% Triton X-100 in PBS for 20 min at room temperature, and blocked for 30 min in PBS containing 0.1% Triton X-100 and 5% normal donkey or goat serum. Primary antibodies included anti-GFAP (rabbit, 1:800; Dako, Z033401-2) and anti-Aβ (mouse monoclonal 82E1, 1:200; IBL, 10323). Sections were incubated overnight at 4 °C, washed in PBS, and then incubated for 1 h at room temperature with Alexa Fluor Plus 594 donkey anti-mouse IgG (1:400; Invitrogen, A32744) or donkey anti-rabbit IgG (1:400;Invitrogen , A32790 ). Sections were washed again, mounted with ProLong Glass Antifade Mount (Invitrogen, P36984), and imaged using a Nikon Eclipse E800 microscope.

Preparation of APPNL-G-F mouse brain homogenates

After the brain dissection, the hippocampus and cerebral cortex of the left hemispheres were frozen on dry ice and stored at − 80 °C. The fresh-frozen tissue was homogenised in Tris buffer (20 mM Tris-HCl, 150 mM NaCl, pH 7.4) supplemented with protease and phosphatase inhibitors (Thermo Fisher, 78440). Homogenates were centrifuged at 50,000 rpm for 20 min at 4 °C, and the resulting supernatant was collected as the TBS fraction. The pellet was resuspended in Tris buffer containing 1% Triton X-100, homogenised, and centrifuged under the same conditions to generate the TBST fraction. For each treatment group (PBS and BRICHOS), homogenates from six mice were pooled, total protein concentrations were determined, and aliquots were stored at − 80 °C until use.

Immunoprecipitation (IP) assay

IP was performed using the Dynabeads Protein G Immunoprecipitation Kit (Thermo Fisher, 80105 G). To minimise nonspecific binding, 5 µL of Protein G beads were washed three times with the kit-supplied buffer and incubated with human or mouse brain extracts for 30 min at room temperature. This pre-clearing step was performed without antibody, and the resulting supernatant was collected for immunodepletion. For antibody coupling, 50 µL of Protein G coated Dynabeads were incubated with 5 µg of Aβ42-specific 21F12 antibody or isotype control (Thermo Fisher, 02-6300) for 1 h at room temperature. Beads were then placed on a magnetic rack, and the supernatant was discarded. After a single wash with 200 µL of wash buffer, antibody-bound beads were mixed with the pre-cleared brain extracts and rotated overnight at 4 °C to allow immunoprecipitation. The next day, samples were centrifuged at 6000 rpm for 3 min at 4 °C. Approximately 80% of the supernatant was collected, aliquoted, and stored at − 80 °C in low-binding protein tubes for downstream analysis.

Antibody labelling for SiMPull assay

The 21F12 antibody was conjugated with Biotin or Alexa Fluor 647 using a Lightning-Link® conjugation kit (Abcam) according to the manufacturer’s instructions. In brief, 10 µL of antibody was mixed with 1 µL of modifier reagent and transferred to the lyophilised conjugation mix. The reaction was incubated for 15 min at room temperature in the dark, followed by addition of 1 µL quencher reagent. After a further 5 min incubation, the labelled antibody was stored at 4 °C until use.

Single-molecule pull-down (SiMPull) assays for detecting Aβ42 aggregates in brain extracts

SiMPull assay was performed on covalently PEGylated glass coverslips (VWR, MENZBC026076AC40) as previously described49, with minor modifications. Assay wells were first coated with 10 µL of NeutrAvidin (Thermo Fisher, 31000; 0.2 mg/mL in PBS containing 0.05% Tween-20) for 5 min, washed, and incubated with biotinylated 21F12 antibody (10 nM in PBS) for 15 min. After washing, recombinant protein preparations or human/mouse brain extracts were added and incubated for 1 h at room temperature. Bound Aβ species were detected using either Alexa Fluor 647–labelled 21F12 antibody (1 nM in PBS) or anti-pyroglutamate antibody (Abcam, IBR-10-2-3), incubated for 30 min. Between each step, wells were washed with PBS containing 0.05% and 1% Tween-20. Finally, 3 µL of PBS was added to each well, and samples were sealed with a second plasma-cleaned coverslip.

Super-resolution imaging of Aβ aggregates

Super-resolution imaging of recombinant Aβ42 aggregates was performed using direct stochastic optical reconstruction microscopy (dSTORM) combined with the SiMPull approach. Aβ42 aggregates were first captured on the surface using biotinylated 6E10 antibodies and subsequently detected with Alexa Fluor 647–labelled 6E10 antibodies. Following the imaging antibody incubation step of the SiMPull protocol, the PBS buffer was removed and replaced with a freshly prepared dSTORM imaging buffer containing 1 mg/mL glucose oxidase, 50 µg/mL catalase, and 50 mM cysteamine in 50 mM Tris buffer supplemented with 10% (v/v) glucose in PBS (pH 8), which served as an oxygen-scavenging system. Images were acquired using 647 nm laser excitation with an exposure time of 30 ms, and a total of 3000 frames were recorded for each acquisition. Localisation of individual blinking events was performed using the Peak Fit module of the GDSC Single Molecule Light Microscopy plugin for ImageJ. Localisations were filtered using a signal intensity threshold of 40 a.u. and a localisation precision threshold of 30 nm. Image reconstruction and analysis were carried out as described previously87.

Separation of Aβ42 aggregates by sucrose gradient and ThT-based single-aggregate imaging

Aliquots of 3 µM Aβ42 aggregation reactions were collected at 85 minutes, when both class A and class B species are present. For fractionation, 400 µL of the aggregation mixture was layered onto a sucrose step gradient ranging from 10% to 50% (w/v) in PBS, prepared in 10% increments. The density of the sucrose solutions was verified by refractive index measurements. Gradients were centrifuged for approximately 5 h at ~ 25,000 × g at room temperature. Fractions were collected immediately after centrifugation and used directly for subsequent analyses.

For ThT-based single-aggregate imaging23, fractions corresponding to 10–20% and 30–40% sucrose were mixed separately. Glass coverslips were first plasma-cleaned using an argon plasma cleaner for 20 min and subsequently coated with poly-L-lysine (Merck, P4707) for 20 min. The collected fractions were mixed with 5 µM ThT and added to the coverslips. After a 15 min incubation, imaging was performed using TIRF microscopy.

Surface-exposed hydrophobicity of Aβ aggregates was assessed using 8-anilino-1-naphthalenesulfonic acid (ANS) fluorescence. Aβ aggregate samples were diluted to a final concentration of 2 μM monomer equivalent and incubated with 40 μM ANS in 96-well half-area plates. Fluorescence spectra were recorded using a Clariostar plate reader with excitation at 350 nm and emission collected from 420–600 nm.

Influence of Ca2+ buffer and lipid vesicles on Aβ42 aggregation under assay conditions

To assess whether the Ca2+ imaging buffer used in the membrane permeabilisation assay alters the aggregate population, pre-formed Aβ42 aggregates (3 µM aggregation, collected at 85 min) were transferred into the Ca2+ buffer and incubated for up to 30 min. Aggregate number was quantified at the beginning and end of the incubation using single-aggregate imaging.

To evaluate potential effects of lipid membranes, monomeric Aβ42 or pre-formed aggregates (3 µM aggregation, collected at 85 min) were incubated separately with lipid vesicles in Eppendorf tubes using the same vesicle composition and concentrations as employed in the membrane permeabilisation assay. After incubation, vesicles were lysed with 0.5% Tween-20, and the samples were analysed using the same single-aggregate detection method to quantify aggregate abundance.

Differentiation of NPCs into cortical neurons and its characterisation

Neural progenitor cells (NPCs) and cortical neurons were derived from iPSCs following a published protocol with minor modifications88. The iPSC line KOLF2.1 J (JIPSC001162) was obtained from The Jackson Laboratory and used in accordance with the supplier’s usage terms. This is a control line carrying the APOE3/3 genotype, which is considered neutral with respect to AD risk. Since our aim was to assess the neurotoxic potential of Aβ aggregates, we selected this APOE3/3 control line to generate healthy neurons that provide a neutral background. Cells were thawed, maintained in StemFlex medium (Gibco, A3349401) on Matrigel-coated plates, and passaged using ReLeSR (STEMCELL Technologies, 100-0484). Colonies were regularly examined to confirm typical iPSC morphology, and pluripotency was verified by immunostaining for established markers. Neural induction was initiated when cultures reached ~ 70–80% confluency. iPSCs were dissociated to single cells with Accutase, replated at high density, and cultured in neural induction medium containing dual-SMAD inhibitors SB431542 (STEMCELL Technologies, 72234) and LDN-193189 (Selleckchem, DM3189), alongside a ROCK inhibitor to promote survival. Within the first week, cells formed a confluent neuroepithelial layer, and neural rosettes typically emerged between DIV12-15. Rosettes were manually selected and expanded using the STEMdiff Neuronal Rosette Selection Reagent (STEMCELL Technologies, 05832). Cultures were maintained in neural maintenance medium, with or without bFGF supplementation depending on the stage of rosette development. Rosettes were routinely frozen between DIV35-41 to establish working stocks, and thawed as needed for further differentiation.

For cortical neuron generation, DIV was reset to zero on the day of neuronal replating, Neural rosettes were dissociated at 7 day post treatment  and replated onto poly-L-ornithine/laminin-coated plates in neuronal plating medium supplemented with DAPT and RevitaCell. Neurons were then maintained in Neurobasal Plus medium with B27, GlutaMAX, and antibiotics. Proliferating cells were suppressed by the addition of 5-fluorouracil/uridine (5FU) starting from DIV3, whilst DAPT was maintained for 7 days post-plating. From DIV7, the medium was supplemented with the neurotrophic factors BDNF (Qkine, Qk050-1000) and GDNF (Qkine, Qk051-1000) (10 ng/mL each) to support maturation and survival. Cells were maintained with half-medium changes two to three times per week. Mature neuronal networks were established by DIV28-30.

Cells cultured in 96-well plates were washed with PBS and fixed in 3.7% paraformaldehyde for 20 min at room temperature. Following permeabilisation with 0.2% Triton X-100/PBS for 10 min, nonspecific binding was blocked using 0.2% fish gelatine (Merck, G7765) in PBS for 1 h. Primary antibodies were diluted in blocking solution and incubated with cells overnight at 4 °C: Nanog (Cell Signaling, 4903; 1:500), Pax6 (BioLegend, 901302; 1:500), FoxG1 (Invitrogen, PA5-2679; 1:25), Ctip2 (Abcam, ab18465; 1:300), βIII-tubulin (Abcam, ab18207; 1:1000), and MAP2 (Abcam, ab5392; 1:5000). The following day, cells were washed and incubated for 1 h at room temperature with Alexa Fluor-conjugated secondary antibodies (Invitrogen; 1:500). Nuclei were counterstained with Hoechst 33342 (Life Technologies, H3570, 1:5000) for 10 min. After final washes in PBS, stained cells were imaged using an Opera Phenix™ High Content Imaging System (PerkinElmer).

Live-neuronal imaging for neurite length measurement

To assess the neurotoxic effects of Aβ aggregates, hPSC neurons were plated in 96-well plates at a density of 30,000 cells per well and placed in an IncuCyte live-cell imaging system and monitored for 5 days. The Aβ42 aggregates used for this experiment were collected from the same aggregation reactions used for the membrane permeabilisation assays. On DIV28 and DIV30, neurons were treated with 0.1 µM Aβ42 oligomers (either alone or prepared in the presence of BRICHOS R221E) at 48 h intervals. Cultures were imaged every 6 h at nine  different fields of view per well. Neuronal network integrity was quantified as neurite length using the IncuCyte® NeuroTrack Analysis Module. A schematic of the neuronal culture timeline, quantification of key markers and treatment paradigm for the neurite outgrowth assay is shown in Supplementary Fig. 20.

iAstrocyte culture and response to Arctic Aβ42

iAstrocytes were generated from fibroblasts obtained from a healthy control (female, age 72) via an induced neuronal progenitor cell (iNPC) intermediate, following previously published differentiation protocols89. Fibroblasts were plated at 250,000 cells per well and transduced for 12 h with retroviral vectors encoding OCT3, SOX2, and KLF4. After 48 h, cultures were transitioned into iNPC medium consisting of DMEM/F12 (Thermo Fisher, 11320033), 1% N2 supplement (Thermo Fisher, 17502048), 1% B27 (Thermo Fisher, A3653401), and FGF (40 ng/mL, Peprotech, 100-18B). To promote astrocytic differentiation, iNPCs were seeded onto fibronectin-coated (5 µg/mL; R&D Systems, 1918-FN) 10 cm dishes in DMEM media with high glucose (ThermoFisher, 11965092) supplemented with 10% FBS (Thermo Fisher, 16000044), 0.2% N2 for 7 days and 50 U/mL of Penicillin-Streptomycin (ThermoFisher, 15140122). iAstrocytes were characterised using established markers vimentin, CD44, and EAAT2. Primary antibodies included anti-vimentin (Merck Millipore, AB5733; 1:1000), anti-CD44 (Abcam, ab157107; 1:200), anti-EAAT2 (Abcam, ab41621; 1:250), and GFAP (Agilent Dako, Z0334, 1:200) diluted in PBS with 1% BSA and 0.1 mg/mL horse serum. Secondary antibodies were Alexa Fluor 488 goat anti-chicken IgG (ThermoFisher, A-11039; 1:1000) for vimentin, and Alexa Fluor 488 donkey anti-rabbit IgG (ThermoFisher, A-21206; 1:1000) for CD44, EAAT2 and GFAP. A schematic of the iAstrocyte culture timeline, marker quantification, and treatment paradigm for the Aβ uptake assay and cytokine measurements in conditioned media is shown in Supplementary Fig. 21.

For experiments with Arctic Aβ42, iAstrocytes were plated at 2,500 cells per well in 96-well plates (Greiner Bio-One) and exposed to Arctic Aβ42 aggregates (0.1 µM monomer equivalents). Cells were incubated for 1 h, after which medium was collected, snap-frozen in liquid nitrogen, and stored at − 80 °C for cytokine analysis. Following exposure, cells were fixed with 4% formaldehyde for 10 min and permeabilised with 0.1% Triton X-100 (Sigma, X100-100ML) in PBS-T (0.1% Tween-20 in PBS) for 30 min, followed by two washes in PBS-T. To block nonspecific binding, cells were incubated in PBS-T containing 5% horse serum for 1 h. Primary antibodies against CD44 (Abcam, ab157107) and Aβ (6E10; BioLegend, 803001) were applied in PBS with 1% BSA and incubated overnight at 4 °C. Secondary antibodies Alexa Fluor 488-labelled goat anti-mouse IgG (Thermo Fisher, A-11029) and Alexa Fluor 647-labelled donkey anti-rabbit IgG (ThermoFisher, A-31573) were added at 1:1000 for 1 h, followed by nuclear counterstaining with 2.5 µM Hoechst (Life Technologies, H3570) for 10 min. iAstrocytes were washed three times in PBS, and internalised Aβ42 was imaged using the Opera Phenix High Content Imaging System (PerkinElmer).

ELISA measurements of cytokines in iAstrocytes media

To assay cytokine secretion by iAstrocytes, cell media were collected after incubation of cells with Aβ samples for 1 h and stored at − 80 °C until analysed. IL-1β, IL-6 and TNF-α in the cell media were measured using the Duoset® enzyme-linked immunosorbent assay (R&D Systems DY201, DY206, DY210), according to the manufacturer’s protocol.

Statistical tests

No statistical method was used to pre-determine sample size, and no data were excluded from the analyses. All membrane permeabilisation assays were performed in three independent biological replicates, which were pooled to obtain sufficient event numbers for reliable kinetic fitting. Cellular experiments were carried out in triplicate biological replicates. Statistical significance between two groups was determined using unpaired two-tailed t tests, whilst comparisons across multiple groups were made using one-way ANOVA followed by Tukey’s post hoc test. Comparisons with P  <  0.05 were considered statistically significant and are indicated with asterisks (*P  <  0.05, **P  <  0.01, ***P  <  0.001 and ****P  <  0.0001). All statistical analyses were performed using Origin 9.0 (OriginLab).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Source data

References

1.Hardy JA Higgins GA Alzheimer’s disease: the amyloid cascade hypothesis Science 1992 256 184 185 1566067 10.1126/science.1566067 Hardy, J. A. & Higgins, G. A. Alzheimer’s disease: the amyloid cascade hypothesis. Science256, 184–185 (1992).10.1126/science.1566067 1566067

2.Knowles TPJ Vendruscolo M Dobson CM The amyloid state and its association with protein misfolding diseases Nat. Rev. Mol. Cell Biol. 2014 15 384 396 10.1038/nrm3810 24854788 Knowles, T. P. J., Vendruscolo, M. & Dobson, C. M. The amyloid state and its association with protein misfolding diseases. Nat. Rev. Mol. Cell Biol.15, 384–396 (2014).10.1038/nrm3810 24854788

3.Hardy, J. & Selkoe, D. J. The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics. Science297, 353–356 (2002).10.1126/science.1072994 12130773

4.Therriault J Biomarker-based staging of Alzheimer disease: rationale and clinical applications Nat. Rev. Neurol. 2024 20 232 244 10.1038/s41582-024-00942-2 38429551 Therriault, J. et al. Biomarker-based staging of Alzheimer disease: rationale and clinical applications. Nat. Rev. Neurol.20, 232–244 (2024).10.1038/s41582-024-00942-2 38429551

5.Mormino EC Amyloid accumulation and cognitive decline in clinically normal older individuals: implications for aging and early alzheimer’s disease J. Alzheimer’s Dis. 2018 64 S633 S646 10.3233/JAD-179928 PMC6387885 29782318 Mormino, E. C. et al. Amyloid accumulation and cognitive decline in clinically normal older individuals: implications for aging and early alzheimer’s disease. J. Alzheimer’s Dis.64, S633–S646 (2018).10.3233/JAD-179928 PMC6387885 29782318

6.Arriagada PV Growdon JH Hedley-Whyte ET Hyman BT Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer’s disease Neurology 1992 42 631 1549228 10.1212/WNL.42.3.631 Arriagada, P. V., Growdon, J. H., Hedley-Whyte, E. T. & Hyman, B. T. Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer’s disease. Neurology42, 631 (1992).10.1212/wnl.42.3.631 1549228

7.Jucker M Walker LC Alzheimer’s disease: From immunotherapy to immunoprevention Cell 2023 186 4260 4270 37729908 PMC10578497 10.1016/j.cell.2023.08.021 Jucker, M. & Walker, L. C. Alzheimer’s disease: From immunotherapy to immunoprevention. Cell186, 4260–4270 (2023).10.1016/j.cell.2023.08.021 PMC10578497 37729908

8.Boxer AL Sperling R Accelerating Alzheimer’s therapeutic development: The past and future of clinical trials Cell. 2023 186 4757 4772 37848035 PMC10625460 10.1016/j.cell.2023.09.023 Boxer, A. L. & Sperling, R. Accelerating Alzheimer’s therapeutic development: The past and future of clinical trials. Cell.186, 4757–4772 (2023).10.1016/j.cell.2023.09.023 PMC10625460 37848035

9.Morris GP Clark IA Vissel B Questions concerning the role of amyloid-β in the definition, aetiology and diagnosis of Alzheimer’s disease Acta Neuropathol. 2018 136 663 689 10.1007/s00401-018-1918-8 30349969 PMC6208728 Morris, G. P., Clark, I. A. & Vissel, B. Questions concerning the role of amyloid-β in the definition, aetiology and diagnosis of Alzheimer’s disease. Acta Neuropathol.136, 663–689 (2018).10.1007/s00401-018-1918-8 PMC6208728 30349969

10.Haass C Selkoe D If amyloid drives Alzheimer disease, why have anti-amyloid therapies not yet slowed cognitive decline? PLOS Biol. 2022 20 e3001694 35862308 PMC9302755 10.1371/journal.pbio.3001694 Haass, C. & Selkoe, D. If amyloid drives Alzheimer disease, why have anti-amyloid therapies not yet slowed cognitive decline? PLOS Biol.20, e3001694 (2022).10.1371/journal.pbio.3001694 PMC9302755 35862308

11.Shankar GM Amyloid-β protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory Nat. Med. 2008 14 837 842 10.1038/nm1782 18568035 PMC2772133 Shankar, G. M. et al. Amyloid-β protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nat. Med.14, 837–842 (2008).10.1038/nm1782 PMC2772133 18568035

12.Benilova I Karran E De Strooper B The toxic Aβ oligomer and Alzheimer’s disease: an emperor in need of clothes Nat. Neurosci. 2012 15 349 357 10.1038/nn.3028 22286176 Benilova, I., Karran, E. & De Strooper, B. The toxic Aβ oligomer and Alzheimer’s disease: an emperor in need of clothes. Nat. Neurosci.15, 349–357 (2012).10.1038/nn.3028 22286176

13.Kayed R Common structure of soluble amyloid oligomers implies common mechanism of Pathogenesis Science. 2003 300 486 489 12702875 10.1126/science.1079469 Kayed, R. et al. Common structure of soluble amyloid oligomers implies common mechanism of Pathogenesis. Science.300, 486–489 (2003).10.1126/science.1079469 12702875

14.Cacciaglia R Soluble Aβ pathology predicts neurodegeneration and cognitive decline independently on p-tau in the earliest Alzheimer’s continuum: Evidence across two independent cohorts Alzheimer’s Dement. 2025 21 e14415 10.1002/alz.14415 PMC11848178 39898436 Cacciaglia, R. et al. Soluble Aβ pathology predicts neurodegeneration and cognitive decline independently on p-tau in the earliest Alzheimer’s continuum: Evidence across two independent cohorts. Alzheimer’s Dement.21, e14415 (2025).10.1002/alz.14415 PMC11848178 39898436

15.Tomic JL Pensalfini A Head E Glabe CG Soluble fibrillar oligomer levels are elevated in Alzheimer’s disease brain and correlate with cognitive dysfunction Neurobiol. Dis. 2009 35 352 358 19523517 PMC2725199 10.1016/j.nbd.2009.05.024 Tomic, J. L., Pensalfini, A., Head, E. & Glabe, C. G. Soluble fibrillar oligomer levels are elevated in Alzheimer’s disease brain and correlate with cognitive dysfunction. Neurobiol. Dis.35, 352–358 (2009).10.1016/j.nbd.2009.05.024 PMC2725199 19523517

16.Lue, L. F. et al. Soluble amyloid β peptide concentration as a predictor of synaptic change in Alzheimer’s disease. Am. J. Pathol.155, 853–862 (1999).10.1016/s0002-9440(10)65184-x PMC1866907 10487842

17.Jarrett JT Berger EP Lansbury PTJ The carboxy terminus of the beta amyloid protein is critical for the seeding of amyloid formation: Implications for the pathogenesis of Alzheimer’s disease Biochemistry 1993 32 4693 4697 8490014 10.1021/bi00069a001 Jarrett, J. T., Berger, E. P. & Lansbury, P. T. J. The carboxy terminus of the beta amyloid protein is critical for the seeding of amyloid formation: Implications for the pathogenesis of Alzheimer’s disease. Biochemistry32, 4693–4697 (1993).10.1021/bi00069a001 8490014

18.Gouras GK Intraneuronal Aβ42 Accumulation in Human Brain Am. J. Pathol. 2000 156 15 20 10623648 PMC1868613 10.1016/S0002-9440(10)64700-1 Gouras, G. K. et al. Intraneuronal Aβ42 Accumulation in Human Brain. Am. J. Pathol.156, 15–20 (2000).10.1016/s0002-9440(10)64700-1 PMC1868613 10623648

19.Iwatsubo T Visualization of Aβ42(43) and Aβ40 in senile plaques with end-specific Aβ monoclonals: Evidence that an initially deposited species is Aβ42(43) Neuron 1994 13 45 53 8043280 10.1016/0896-6273(94)90458-8 Iwatsubo, T. et al. Visualization of Aβ42(43) and Aβ40 in senile plaques with end-specific Aβ monoclonals: Evidence that an initially deposited species is Aβ42(43). Neuron13, 45–53 (1994).10.1016/0896-6273(94)90458-8 8043280

20.De S Different soluble aggregates of Aβ42 can give rise to cellular toxicity through different mechanisms Nat. Commun. 2019 10 1541 10.1038/s41467-019-09477-3 30948723 PMC6449370 De, S. et al. Different soluble aggregates of Aβ42 can give rise to cellular toxicity through different mechanisms. Nat. Commun.10, 1541 (2019).10.1038/s41467-019-09477-3 PMC6449370 30948723

21.Drummond E Wisniewski T Alzheimer’s disease: experimental models and reality Acta Neuropathol. 2017 133 155 175 10.1007/s00401-016-1662-x 28025715 PMC5253109 Drummond, E. & Wisniewski, T. Alzheimer’s disease: experimental models and reality. Acta Neuropathol.133, 155–175 (2017).10.1007/s00401-016-1662-x PMC5253109 28025715

22.Manabe T Saito T A decade progress in the phenotyping of AppNL-G-F knock-in mouse model of Alzheimer’s disease Neurosci. Res. 2025 220 104959 40998178 10.1016/j.neures.2025.104959 Manabe, T. & Saito, T. A decade progress in the phenotyping of AppNL-G-F knock-in mouse model of Alzheimer’s disease. Neurosci. Res.220, 104959 (2025).10.1016/j.neures.2025.104959 40998178

23.De S Klenerman D Imaging individual protein aggregates to follow aggregation and determine the role of aggregates in neurodegenerative disease Biochim. Biophys. Acta Proteins Proteomics 2019 1867 870 878 30611780 PMC6676340 10.1016/j.bbapap.2018.12.010 De, S. & Klenerman, D. Imaging individual protein aggregates to follow aggregation and determine the role of aggregates in neurodegenerative disease. Biochim. Biophys. Acta Proteins Proteomics1867, 870–878 (2019).10.1016/j.bbapap.2018.12.010 PMC6676340 30611780

24.De S Soluble aggregates present in cerebrospinal fluid change in size and mechanism of toxicity during Alzheimer’s disease progression Acta Neuropathol. Commun. 2019 7 1 10.1186/s40478-019-0777-4 PMC6659275 31349874 De, S. et al. Soluble aggregates present in cerebrospinal fluid change in size and mechanism of toxicity during Alzheimer’s disease progression. Acta Neuropathol. Commun.7, 1 (2019).10.1186/s40478-019-0777-4 PMC6659275 31349874

25.Stern AM Abundant Aβ fibrils in ultracentrifugal supernatants of aqueous extracts from Alzheimer’s disease brains Neuron 2023 111 2012 2020.e4 37167969 PMC10330525 10.1016/j.neuron.2023.04.007 Stern, A. M. et al. Abundant Aβ fibrils in ultracentrifugal supernatants of aqueous extracts from Alzheimer’s disease brains. Neuron111, 2012–2020.e4 (2023).10.1016/j.neuron.2023.04.007 PMC10330525 37167969

26.Liang R Amyloid-β oligomers, curvilinear and annular assemblies, imaged by cryo-ET, cryo-EM, and AFM Sci. Adv. 2025 11 eadx9030 40864693 PMC12383235 10.1126/sciadv.adx9030 Liang, R. et al. Amyloid-β oligomers, curvilinear and annular assemblies, imaged by cryo-ET, cryo-EM, and AFM. Sci. Adv.11, eadx9030 (2025).10.1126/sciadv.adx9030 PMC12383235 40864693

27.Ciudad S Aβ(1-42) tetramer and octamer structures reveal edge conductivity pores as a mechanism for membrane damage Nat. Commun. 2020 11 3014 10.1038/s41467-020-16566-1 32541820 PMC7296003 Ciudad, S. et al. Aβ(1-42) tetramer and octamer structures reveal edge conductivity pores as a mechanism for membrane damage. Nat. Commun.11, 3014 (2020).10.1038/s41467-020-16566-1 PMC7296003 32541820

28.Xiao, Y. et al. Aβ(1–42) fibril structure illuminates self-recognition and replication of amyloid in Alzheimer’s disease. Nat. Struct. Mol. Biol. 22, 499–505 (2015).10.1038/nsmb.2991 PMC4476499 25938662

29.Sideris DI Soluble amyloid beta-containing aggregates are present throughout the brain at early stages of Alzheimer’s disease Brain Commun. 2021 3 fcab147 34396107 PMC8361392 10.1093/braincomms/fcab147 Sideris, D. I. et al. Soluble amyloid beta-containing aggregates are present throughout the brain at early stages of Alzheimer’s disease. Brain Commun.3, fcab147 (2021).10.1093/braincomms/fcab147 PMC8361392 34396107

30.Lazar T Connor A DeLisle CF Burger V Tompa P Targeting protein disorder: the next hurdle in drug discovery Nat. Rev. Drug Discov. 2025 24 743 763 10.1038/s41573-025-01220-6 40490488 Lazar, T., Connor, A., DeLisle, C. F., Burger, V. & Tompa, P. Targeting protein disorder: the next hurdle in drug discovery. Nat. Rev. Drug Discov.24, 743–763 (2025).10.1038/s41573-025-01220-6 40490488

31.Khachaturian, Z. S. Calcium hypothesis of Alzheimer’s disease and brain aging. Ann. N. Y. Acad. Sci.747, 1–11 (1994).10.1111/j.1749-6632.1994.tb44398.x 7847664

32.Disterhoft, J. F., Moyer, J. R. & Thompson, L. T. The calcium rationale in aging and Alzheimer’s disease. Evidence from an animal model of normal aging. Ann. N. Y. Acad. Sci.747, 382–406 (1994).10.1111/j.1749-6632.1994.tb44424.x 7847686

33.Arbel-Ornath M Soluble oligomeric amyloid-β induces calcium dyshomeostasis that precedes synapse loss in the living mouse brain Mol. Neurodegener. 2017 12 27 28327181 PMC5361864 10.1186/s13024-017-0169-9 Arbel-Ornath, M. et al. Soluble oligomeric amyloid-β induces calcium dyshomeostasis that precedes synapse loss in the living mouse brain. Mol. Neurodegener.12, 27 (2017).10.1186/s13024-017-0169-9 PMC5361864 28327181

34.Zott B A vicious cycle of β amyloid–dependent neuronal hyperactivation Science 2019 365 559 565 31395777 PMC6690382 10.1126/science.aay0198 Zott, B. et al. A vicious cycle of β amyloid–dependent neuronal hyperactivation. Science365, 559–565 (2019).10.1126/science.aay0198 PMC6690382 31395777

35.Malleret, G. et al. Inducible and reversible enhancement of learning, memory, and long-term potentiation by genetic inhibition of calcineurin. Cell104, 675–686 (2001).10.1016/s0092-8674(01)00264-1 11257222

36.Reese, L. C., Laezza, F., Woltjer, R. & Taglialatela, G. Dysregulated phosphorylation of Ca(2+)/calmodulin-dependent protein kinase II-alpha in the hippocampus of subjects with mild cognitive impairment and Alzheimer’s disease. J. Neurochem.119, 791–804 (2011).10.1111/j.1471-4159.2011.07447.x PMC4021864 21883216

37.Lohse B Bolinger PY Stamou D Encapsulation efficiency measured on single small unilamellar vesicles J. Am. Chem. Soc. 2008 130 14372 14373 18842043 10.1021/ja805030w Lohse, B., Bolinger, P. Y. & Stamou, D. Encapsulation efficiency measured on single small unilamellar vesicles. J. Am. Chem. Soc.130, 14372–14373 (2008).10.1021/ja805030w 18842043

38.Demuro A Smith M Parker I Single-channel Ca2+ imaging implicates Aβ1–42 amyloid pores in Alzheimer’s disease pathology J. Cell Biol. 2011 195 515 524 22024165 PMC3206345 10.1083/jcb.201104133 Demuro, A., Smith, M. & Parker, I. Single-channel Ca2+ imaging implicates Aβ1–42 amyloid pores in Alzheimer’s disease pathology. J. Cell Biol.195, 515–524 (2011).10.1083/jcb.201104133 PMC3206345 22024165

39.Quist A Amyloid ion channels: A common structural link for protein-misfolding disease Proc. Natl. Acad. Sci. USA. 2005 102 10427 10432 16020533 PMC1180768 10.1073/pnas.0502066102 Quist, A. et al. Amyloid ion channels: A common structural link for protein-misfolding disease. Proc. Natl. Acad. Sci. USA.102, 10427–10432 (2005).10.1073/pnas.0502066102 PMC1180768 16020533

40.Chen G Augmentation of Bri2 molecular chaperone activity against amyloid-β reduces neurotoxicity in mouse hippocampus in vitro Commun. Biol. 2020 3 32 10.1038/s42003-020-0757-z 31959875 PMC6971075 Chen, G. et al. Augmentation of Bri2 molecular chaperone activity against amyloid-β reduces neurotoxicity in mouse hippocampus in vitro. Commun. Biol.3, 32 (2020).10.1038/s42003-020-0757-z PMC6971075 31959875

41.Chen G Bri2 BRICHOS client specificity and chaperone activity are governed by assembly state Nat. Commun. 2017 8 2081 10.1038/s41467-017-02056-4 29234026 PMC5727130 Chen, G. et al. Bri2 BRICHOS client specificity and chaperone activity are governed by assembly state. Nat. Commun.8, 2081 (2017).10.1038/s41467-017-02056-4 PMC5727130 29234026

42.Wang Z Learnings about Aβ from human brain recommend the use of a live-neuron bioassay for the discovery of next generation Alzheimer’s disease immunotherapeutics Acta Neuropathol. Commun. 2023 11 39 10.1186/s40478-023-01511-2 36899414 PMC10007750 Wang, Z. et al. Learnings about Aβ from human brain recommend the use of a live-neuron bioassay for the discovery of next generation Alzheimer’s disease immunotherapeutics. Acta Neuropathol. Commun.11, 39 (2023).10.1186/s40478-023-01511-2 PMC10007750 36899414

43.Cohen SIA Proliferation of amyloid-β42 aggregates occurs through a secondary nucleation mechanism Proc. Natl. Acad. Sci. USA. 2013 110 9758 9763 23703910 PMC3683769 10.1073/pnas.1218402110 Cohen, S. I. A. et al. Proliferation of amyloid-β42 aggregates occurs through a secondary nucleation mechanism. Proc. Natl. Acad. Sci. USA.110, 9758–9763 (2013).10.1073/pnas.1218402110 PMC3683769 23703910

44.Michaels TCT Dynamics of oligomer populations formed during the aggregation of Alzheimer’s Aβ42 peptide Nat. Chem. 2020 12 445 451 10.1038/s41557-020-0452-1 32284577 PMC7616962 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).10.1038/s41557-020-0452-1 PMC7616962 32284577

45.Dear AJ Identification of on- and off-pathway oligomers in amyloid fibril formation Chem. Sci. 2020 11 6236 6247 32953019 PMC7480182 10.1039/C9SC06501F Dear, A. J. et al. Identification of on- and off-pathway oligomers in amyloid fibril formation. Chem. Sci.11, 6236–6247 (2020).10.1039/c9sc06501f PMC7480182 32953019

46.Dear AJ Kinetic diversity of amyloid oligomers Proc. Natl. Acad. Sci. USA. 2020 117 12087 12094 32414930 PMC7275774 10.1073/pnas.1922267117 Dear, A. J. et al. Kinetic diversity of amyloid oligomers. Proc. Natl. Acad. Sci. USA.117, 12087–12094 (2020).10.1073/pnas.1922267117 PMC7275774 32414930

47.Cohen SIA A molecular chaperone breaks the catalytic cycle that generates toxic Aβ oligomers Nat. Struct. Mol. Biol. 2015 22 207 213 10.1038/nsmb.2971 25686087 PMC4595974 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).10.1038/nsmb.2971 PMC4595974 25686087

48.Hong W Diffusible, highly bioactive oligomers represent a critical minority of soluble Aβ in Alzheimer’s disease brain Acta Neuropathol. 2018 136 19 40 10.1007/s00401-018-1846-7 29687257 PMC6647843 Hong, W. et al. Diffusible, highly bioactive oligomers represent a critical minority of soluble Aβ in Alzheimer’s disease brain. Acta Neuropathol.136, 19–40 (2018).10.1007/s00401-018-1846-7 PMC6647843 29687257

49.Xia Z Co-aggregation with Apolipoprotein E modulates the function of Amyloid-β in Alzheimer’s disease Nat. Commun. 2024 15 4695 10.1038/s41467-024-49028-z 38824138 PMC11144216 Xia, Z. et al. Co-aggregation with Apolipoprotein E modulates the function of Amyloid-β in Alzheimer’s disease. Nat. Commun.15, 4695 (2024).10.1038/s41467-024-49028-z PMC11144216 38824138

50.Flagmeier, P. et al. Direct measurement of lipid membrane disruption connects kinetics and toxicity of Aβ42 aggregation. Nat. Struct. Mol. Biol.27, 886–891 (2020).10.1038/s41594-020-0471-z 32778821

51.Saito, T. et al. Single App knock-in mouse models of Alzheimer’s disease. Nat. Neurosci.17, 661 (2014).10.1038/nn.3697 24728269

52.Manchanda S Intravenous treatment with a molecular chaperone designed against β-amyloid toxicity improves Alzheimer’s disease pathology in mouse models Mol. Ther. 2023 31 487 502 35982621 10.1016/j.ymthe.2022.08.010 PMC9931549 Manchanda, S. et al. Intravenous treatment with a molecular chaperone designed against β-amyloid toxicity improves Alzheimer’s disease pathology in mouse models. Mol. Ther.31, 487–502 (2023).10.1016/j.ymthe.2022.08.010 PMC9931549 35982621

53.Zhang R Intravenous chaperone treatment of late-stage Alzheimer´s disease (AD) mouse model affects amyloid plaque load, reactive gliosis and AD-related genes Transl. Psychiatry 2024 14 453 10.1038/s41398-024-03161-x 39448576 PMC11502864 Zhang, R. et al. Intravenous chaperone treatment of late-stage Alzheimer´s disease (AD) mouse model affects amyloid plaque load, reactive gliosis and AD-related genes. Transl. Psychiatry14, 453 (2024).10.1038/s41398-024-03161-x PMC11502864 39448576

54.Poojari, C. & Strodel, B. Stability of transmembrane amyloid β-peptide and membrane integrity tested by molecular modeling of site-specific Aβ42 Mutations. PLoS ONE8, e78399 (2013).10.1371/journal.pone.0078399 PMC3820573 24244308

55.Koistinaho M Apolipoprotein E promotes astrocyte colocalization and degradation of deposited amyloid-β peptides Nat. Med. 2004 10 719 726 10.1038/nm1058 15195085 Koistinaho, M. et al. Apolipoprotein E promotes astrocyte colocalization and degradation of deposited amyloid-β peptides. Nat. Med.10, 719–726 (2004).10.1038/nm1058 15195085

56.Ashe KH The biogenesis and biology of amyloid β oligomers in the brain Alzheimer’s Dement. 2020 16 1561 1567 10.1002/alz.12084 PMC7984270 32543725 Ashe, K. H. The biogenesis and biology of amyloid β oligomers in the brain. Alzheimer’s Dement.16, 1561–1567 (2020).10.1002/alz.12084 PMC7984270 32543725

57.Sakono M Zako T Amyloid oligomers: formation and toxicity of Aβ oligomers FEBS J. 2010 277 1348 1358 20148964 10.1111/j.1742-4658.2010.07568.x Sakono, M. & Zako, T. Amyloid oligomers: formation and toxicity of Aβ oligomers. FEBS J.277, 1348–1358 (2010).10.1111/j.1742-4658.2010.07568.x 20148964

58.Barghorn S Globular amyloid β-peptide1−42 oligomer − a homogenous and stable neuropathological protein in Alzheimer’s disease J. Neurochem. 2005 95 834 847 16135089 10.1111/j.1471-4159.2005.03407.x Barghorn, S. et al. Globular amyloid β-peptide1−42 oligomer − a homogenous and stable neuropathological protein in Alzheimer’s disease. J. Neurochem.95, 834–847 (2005).10.1111/j.1471-4159.2005.03407.x 16135089

59.Noguchi, A. et al. Isolation and Characterization of Patient-derived, Toxic, High Mass Amyloid β-Protein (Aβ) Assembly from Alzheimer Disease Brains. J. Biol. Chem.284, 32895–32905 (2009).10.1074/jbc.M109.000208 PMC2781705 19759000

60.Tang H Emerging biophysical origins and pathogenic implications of amyloid oligomers Nat. Commun. 2025 16 2937 10.1038/s41467-025-58335-y 40133283 PMC11937510 Tang, H. et al. Emerging biophysical origins and pathogenic implications of amyloid oligomers. Nat. Commun.16, 2937 (2025).10.1038/s41467-025-58335-y PMC11937510 40133283

61.Curk S Self-replication of Aβ42 aggregates occurs on small and isolated fibril sites Proc. Natl. Acad. Sci. USA 2024 121 e2220075121 38335256 PMC10873593 10.1073/pnas.2220075121 Curk, S. et al. Self-replication of Aβ42 aggregates occurs on small and isolated fibril sites. Proc. Natl. Acad. Sci. USA121, e2220075121 (2024).10.1073/pnas.2220075121 PMC10873593 38335256

62.Nirmalraj, P. N., Bhattacharya, S. & Thompson, D. Accelerated Alzheimer’s Aβ-42 secondary nucleation chronologically visualized on fibril surfaces. Sci. Adv.10, eadp5059 (2024).10.1126/sciadv.adp5059 PMC11506133 39454002

63.Hu J Structural defects in amyloid-β fibrils drive secondary nucleation Nat. Commun. 2026 17 1933 10.1038/s41467-026-69377-1 41708600 PMC12923811 Hu, J. et al. Structural defects in amyloid-β fibrils drive secondary nucleation. Nat. Commun.17, 1933 (2026).10.1038/s41467-026-69377-1 PMC12923811 41708600

64.Adam L Specific inhibition of α-synuclein oligomer generation and toxicity by the chaperone domain Bri2 BRICHOS Protein Sci. 2024 33 e5091 38980078 PMC11232276 10.1002/pro.5091 Adam, L. et al. Specific inhibition of α-synuclein oligomer generation and toxicity by the chaperone domain Bri2 BRICHOS. Protein Sci.33, e5091 (2024).10.1002/pro.5091 PMC11232276 38980078

65.Leppert A Smallest Secondary Nucleation Competent Aβ Aggregates Probed by an ATP-Independent Molecular Chaperone Domain Biochemistry 2021 60 678 688 33621049 PMC8028046 10.1021/acs.biochem.1c00003 Leppert, A. et al. Smallest Secondary Nucleation Competent Aβ Aggregates Probed by an ATP-Independent Molecular Chaperone Domain. Biochemistry60, 678–688 (2021).10.1021/acs.biochem.1c00003 PMC8028046 33621049

66.Ladiwala ARA Conformational Differences between Two Amyloid β Oligomers of Similar Size and Dissimilar Toxicity* J. Biol. Chem. 2012 287 24765 24773 22547072 PMC3397903 10.1074/jbc.M111.329763 Ladiwala, A. R. A. et al. Conformational Differences between Two Amyloid β Oligomers of Similar Size and Dissimilar Toxicity*. J. Biol. Chem.287, 24765–24773 (2012).10.1074/jbc.M111.329763 PMC3397903 22547072

67.Rinauro DJ Chiti F Vendruscolo M Limbocker R Misfolded protein oligomers: mechanisms of formation, cytotoxic effects, and pharmacological approaches against protein misfolding diseases Mol. Neurodegener. 2024 19 20 38378578 PMC10877934 10.1186/s13024-023-00651-2 Rinauro, D. J., Chiti, F., Vendruscolo, M. & Limbocker, R. Misfolded protein oligomers: mechanisms of formation, cytotoxic effects, and pharmacological approaches against protein misfolding diseases. Mol. Neurodegener.19, 20 (2024).10.1186/s13024-023-00651-2 PMC10877934 38378578

68.Schützmann MP Hoyer W Off-pathway oligomers of α-synuclein and Aβ inhibit secondary nucleation of α-synuclein amyloid fibrils J. Mol. Biol. 2025 437 169048 40015369 10.1016/j.jmb.2025.169048 Schützmann, M. P. & Hoyer, W. Off-pathway oligomers of α-synuclein and Aβ inhibit secondary nucleation of α-synuclein amyloid fibrils. J. Mol. Biol.437, 169048 (2025).10.1016/j.jmb.2025.169048 40015369

69.Brody DL Jiang H Wildburger N Esparza TJ Non-canonical soluble amyloid-beta aggregates and plaque buffering: controversies and future directions for target discovery in Alzheimer’s disease Alzheimers. Res. Ther. 2017 9 62 28818091 PMC5561579 10.1186/s13195-017-0293-3 Brody, D. L., Jiang, H., Wildburger, N. & Esparza, T. J. Non-canonical soluble amyloid-beta aggregates and plaque buffering: controversies and future directions for target discovery in Alzheimer’s disease. Alzheimers. Res. Ther.9, 62 (2017).10.1186/s13195-017-0293-3 PMC5561579 28818091

70.Abelein A Johansson J Amyloid inhibition by molecular chaperones in vitro can be translated to Alzheimer’s pathology in vivo RSC Med. Chem. 2023 14 848 857 37252101 PMC10211315 10.1039/D3MD00040K Abelein, A. & Johansson, J. Amyloid inhibition by molecular chaperones in vitro can be translated to Alzheimer’s pathology in vivo. RSC Med. Chem.14, 848–857 (2023).10.1039/d3md00040k PMC10211315 37252101

71.Glass CK Saijo K Winner B Marchetto MC Gage FH Mechanisms underlying inflammation in neurodegeneration Cell 2010 140 918 934 20303880 PMC2873093 10.1016/j.cell.2010.02.016 Glass, C. K., Saijo, K., Winner, B., Marchetto, M. C. & Gage, F. H. Mechanisms underlying inflammation in neurodegeneration. Cell140, 918–934 (2010).10.1016/j.cell.2010.02.016 PMC2873093 20303880

72.Saijo K A Nurr1/CoREST pathway in microglia and astrocytes protectsdopaminergic neurons from inflammation-induced Death Cell 2009 137 47 59 19345186 PMC2754279 10.1016/j.cell.2009.01.038 Saijo, K. et al. A Nurr1/CoREST pathway in microglia and astrocytes protectsdopaminergic neurons from inflammation-induced Death. Cell137, 47–59 (2009).10.1016/j.cell.2009.01.038 PMC2754279 19345186

73.Urbanek, A. et al. Molecular determinants of protein pathogenicity at the single-aggregate level. Adv. Sci. 12, 24010229 (2025).10.1002/advs.202410229 PMC11884545 39804980

74.Wang W-Y Tan M-S Yu J-T Tan L Role of pro-inflammatory cytokines released from microglia in Alzheimer’s disease Ann. Transl. Med. 2015 3 136 26207229 PMC4486922 10.3978/j.issn.2305-5839.2015.03.49 Wang, W.-Y., Tan, M.-S., Yu, J.-T. & Tan, L. Role of pro-inflammatory cytokines released from microglia in Alzheimer’s disease. Ann. Transl. Med.3, 136 (2015).10.3978/j.issn.2305-5839.2015.03.49 PMC4486922 26207229

75.Whiten, D. R. et al. Tumour necrosis factor induces increased production of extracellular amyloid-β- and α-synuclein-containing aggregates by human Alzheimer’s disease neurons. Brain Commun. 2, fcaa146, (2020)10.1093/braincomms/fcaa146 PMC7850285 33543132

76.Ringheim GE Enhancement of beta-amyloid precursor protein transcription and expression by the soluble interleukin-6 receptor/interleukin-6 complex Mol. Brain Res. 1998 55 35 44 9645958 10.1016/S0169-328X(97)00356-2 Ringheim, G. E. et al. Enhancement of beta-amyloid precursor protein transcription and expression by the soluble interleukin-6 receptor/interleukin-6 complex. Mol. Brain Res.55, 35–44 (1998).10.1016/s0169-328x(97)00356-2 9645958

77.Bouter Y N-truncated amyloid β (Aβ) 4-42 forms stable aggregates and induces acute and long-lasting behavioral deficits Acta Neuropathol. 2013 126 189 205 10.1007/s00401-013-1129-2 23685882 PMC3722453 Bouter, Y. et al. N-truncated amyloid β (Aβ) 4-42 forms stable aggregates and induces acute and long-lasting behavioral deficits. Acta Neuropathol.126, 189–205 (2013).10.1007/s00401-013-1129-2 PMC3722453 23685882

78.Murphy MP LeVine III H Alzheimer’s Disease and the Amyloid-β Peptide J. Alzheimer’s Dis. 2010 19 311 323 10.3233/JAD-2010-1221 PMC2813509 20061647 Murphy, M. P. & LeVine III, H. Alzheimer’s Disease and the Amyloid-β Peptide. J. Alzheimer’s Dis.19, 311–323 (2010).10.3233/JAD-2010-1221 PMC2813509 20061647

79.Hampel H The Amyloid-β Pathway in Alzheimer’s Disease Mol. Psychiatry 2021 26 5481 5503 10.1038/s41380-021-01249-0 34456336 PMC8758495 Hampel, H. et al. The Amyloid-β Pathway in Alzheimer’s Disease. Mol. Psychiatry26, 5481–5503 (2021).10.1038/s41380-021-01249-0 PMC8758495 34456336

80.Hellstrand E Boland B Walsh DM Linse S Amyloid β-protein aggregation produces highly reproducible kinetic data and occurs by a two-phase process ACS Chem. Neurosci. 2010 1 13 18 22778803 10.1021/cn900015v PMC3368626 Hellstrand, E., Boland, B., Walsh, D. M. & Linse, S. Amyloid β-protein aggregation produces highly reproducible kinetic data and occurs by a two-phase process. ACS Chem. Neurosci.1, 13–18 (2010).10.1021/cn900015v PMC3368626 22778803

81.Brezinova M Identification of potent high-affinity secondary nucleation inhibitors of Aβ42 aggregation from an ultra-large chemical library using deep docking Mol. Syst. Biol. 2026 22 69 87 10.1038/s44320-025-00159-5 41193694 PMC12759071 Brezinova, M. et al. Identification of potent high-affinity secondary nucleation inhibitors of Aβ42 aggregation from an ultra-large chemical library using deep docking. Mol. Syst. Biol.22, 69–87 (2026).10.1038/s44320-025-00159-5 PMC12759071 41193694

82.Flagmeier P Ultrasensitive measurement of Ca2+ influx into lipid vesicles induced by protein aggregates Angew. Chemie Int. Ed. 2017 56 7750 7754 10.1002/anie.201700966 PMC5615231 28474754 Flagmeier, P. et al. Ultrasensitive measurement of Ca2+ influx into lipid vesicles induced by protein aggregates. Angew. Chemie Int. Ed.56, 7750–7754 (2017).10.1002/anie.201700966 PMC5615231 28474754

83.Drews A Inhibiting the Ca2+ influx induced by human CSF Cell Rep. 2017 21 3310 3316 29241555 PMC5745229 10.1016/j.celrep.2017.11.057 Drews, A. et al. Inhibiting the Ca2+ influx induced by human CSF. Cell Rep.21, 3310–3316 (2017).10.1016/j.celrep.2017.11.057 PMC5745229 29241555

84.Lee J Amyloid β Ion Channels in a Membrane Comprising Brain Total Lipid Extracts ACS Chem. Neurosci. 2017 8 1348 1357 28135799 PMC6197823 10.1021/acschemneuro.7b00006 Lee, J. et al. Amyloid β Ion Channels in a Membrane Comprising Brain Total Lipid Extracts. ACS Chem. Neurosci.8, 1348–1357 (2017).10.1021/acschemneuro.7b00006 PMC6197823 28135799

85.Dear AJ The catalytic nature of protein aggregation J. Chem. Phys. 2020 152 45101 10.1063/1.5133635 PMC7377910 32007046 Dear, A. J. et al. The catalytic nature of protein aggregation. J. Chem. Phys.152, 45101 (2020).10.1063/1.5133635 PMC7377910 32007046

86.Meisl G Molecular mechanisms of protein aggregation from global fitting of kinetic models Nat. Protoc. 2016 11 252 272 10.1038/nprot.2016.010 26741409 Meisl, G. et al. Molecular mechanisms of protein aggregation from global fitting of kinetic models. Nat. Protoc.11, 252–272 (2016).10.1038/nprot.2016.010 26741409

87.Cox D Quantitative profiling of nanoscopic protein aggregates reveals specific fingerprint of TDP-43-positive assemblies in motor neuron disease Adv. Sci. 2025 12 e05484 10.1002/advs.202505484 PMC12667529 40985163 Cox, D. et al. Quantitative profiling of nanoscopic protein aggregates reveals specific fingerprint of TDP-43-positive assemblies in motor neuron disease. Adv. Sci.12, e05484 (2025).10.1002/advs.202505484 PMC12667529 40985163

88.Dannert A Klimmt J Cardoso Gonçalves C Crusius D Paquet D Reproducible and scalable differentiation of highly pure cortical neurons from human induced pluripotent stem cells STAR Protoc. 2023 4 102266 37148244 PMC10193006 10.1016/j.xpro.2023.102266 Dannert, A., Klimmt, J., Cardoso Gonçalves, C., Crusius, D. & Paquet, D. Reproducible and scalable differentiation of highly pure cortical neurons from human induced pluripotent stem cells. STAR Protoc.4, 102266 (2023).10.1016/j.xpro.2023.102266 PMC10193006 37148244

89.Bell SM Increasing hexokinase 1 expression improves mitochondrial and glycolytic functional deficits seen in sporadic Alzheimer’s disease astrocytes Mol. Psychiatry 2025 30 1369 1382 10.1038/s41380-024-02746-8 39271753 PMC11919762 Bell, S. M. et al. Increasing hexokinase 1 expression improves mitochondrial and glycolytic functional deficits seen in sporadic Alzheimer’s disease astrocytes. Mol. Psychiatry30, 1369–1382 (2025).10.1038/s41380-024-02746-8 PMC11919762 39271753

Come citare questo studio

Agnieszka Urbanek, Alexander J Dear, David P Rhodes, Emma F Garland, Hollie E Wareing, Harry J Baker, Willem Molenkamp, Ruixin Zhang, Gefei Chen, Axel Abelein, Simon M Bell, Juan A Varela, Thomas C T Michaels, J Robin Highley, Georg Meisl, Jan Johansson, Suman De. (2026). Two functionally distinct amyloid-β intermediates link aggregation pathways to pathology in Alzheimer's disease.. Nature communications. DOI: 10.1038/s41467-026-79028-0

Consulta la pubblicazione originale ↗