Diverse Structural Conversion and Aggregation Pathways of Alzheimer’s Amyloid-β (1–40)

Abstract

Complex amyloid aggregation of amyloid-β (1–40) (Aβ_1–40) in terms of monomer structures has not been fully understood. Herein, we report the microscopic mechanism and pathways of Aβ_1–40 aggregation with macroscopic viewpoints through tuning its initial structure and solubility. Partial helical structures of Aβ_1–40 induced by low solvent polarity accelerated cytotoxic Aβ_1–40 amyloid fibrillation, while predominantly helical folds did not aggregate. Changes in the solvent polarity caused a rapid formation of β-structure-rich protofibrils or oligomers via aggregation-prone helical structures. Modulation of the pH and salt concentration transformed oligomers to protofibrils, which proceeded to amyloid formation. We reveal diverse molecular mechanisms underlying Aβ_1–40 aggregation with conceptual energy diagrams and propose that aggregation-prone partial helical structures are key to inducing amyloidogenesis. We demonstrate that context-dependent protein aggregation is comprehensively understood using the macroscopic phase diagram, which provides general insights into differentiation of amyloid formation and phase separation from unfolded and folded structures.

Graphical Abstract

Destabilized soluble proteins misfold and self-assemble to insoluble aggregates, which triggers a loss of biological function and progression of disease. In the last two decades, protein misfolding and aggregation have attracted considerable attention in many research fields. A number of studies have been conducted to elucidate the role of protein aggregation in the pathogenesis of more than 50 disorders, including diverse neurodegenerative diseases and systematic amyloidosis, as well as in homeostasis in various cells. The biologically beneficial functions of protein aggregates have also been revealed in fungal coat formation, biofilm formation, melanin biosynthesis, and the immune system.^1,2 Innovative attempts to design and generate protein aggregate-based biomaterials have recently increased in terms of biotechnology.^3,4 Phase separation is one type of protein aggregation for biological functions and has garnered significant interest in recent years.⁵

Proteins have shown the capability to form various types of aggregates, including oligomers, amorphous aggregates, and fibrillar aggregates such as immature and mature amyloid fibrils, depending on the conditions. Significant efforts have been made to characterize the structural and morphological properties of individual aggregates at atomic and molecular levels.^6–9 Oligomers and amorphous aggregates generally show spherical morphologies with or without defined secondary structures, while immature and mature amyloid fibrils exhibit fibrillar morphologies with a β-structure-rich core and unstructured flanking regions.⁸

Among various protein aggregates, the formation of amyloid fibrils and their structures have been well examined. Amyloid fibrils exhibit hierarchical suprastructures. They consist of bundles of several protofilaments that are further composed of β-sheets with hydrogen bond networks among monomers.¹ Although the basic mechanisms responsible for amyloid formation have been uncovered, obtaining more detailed information on the sophisticated mechanisms underlying amyloidogenesis remains elusive. Amyloid fibrillation has shown crystal-like formation. Productive nucleation limits the entire process by introducing a lag time, and a rapid fibril growth phase subsequently emerges by consuming precursor proteins.¹⁰ Sonication has shown to be highly effective in inducing amyloid formation in vitro.^11–13 Effects of sonication are rationalized by the generation and burst of cavitation microbubbles. Bubbles are created and collapsed in our body during or after decompression from pressure exposures such as those undergone by scuba divers, astronauts, and caisson and tunnel workers.¹⁴ Cavitation bubbles provide air (hydrophobic)–water (hydrophilic) interfaces for condensing precursor proteins to promote aggregation. Of note, large mechanical forces have been shown to cause the formation of cavitation bubbles in the brain, which may be responsible for Alzheimer’s disease.¹⁵ Burst of bubbles is accompanied by large shearing forces as well as high temperature and pressure, which may be efficient for the primary nucleation and secondary nucleation pathway. Physiological fluids at the interface between hydrophilic and hydrophobic surfaces generate shearing forces.¹⁶ Thus, sonication can mimic physiological and pathological conditions for aggregation-related diseases.

Pre-amyloid aggregates are often detected during amyloidogenesis. Small oligomers of amyloid-β (Aβ),¹⁷ α-synuclein (αSN),¹⁸ and human islet amyloid polypeptide (hIAPP)¹⁹ grow to larger oligomers that are ultimately converted to amyloid fibrils. The self-assembly of immature fibrils of Aβ,²⁰ polyglutamine,²¹ and tau,²² i.e., protofibrils, precedes amyloid fibrillation. These nonmature amyloid aggregates have been identified under certain conditions as end products that do not proceed to amyloid fibrillation;²³ therefore, these aggregated species have been characterized as off-pathway intermediates.

Biophysical studies on the aggregation-competent structures of precursor proteins at the atomic and secondary structure levels have markedly increased our understanding of the mechanisms underlying amyloid fibrillation. Native-like intermediates of the SH3 domain with an extended C-terminus showed higher amyloidogenicity than native-state structures.²⁴ Native-like β2-microglobulin (β2m) intermediates with transPro 32 and frataxin in the folding transition state have been suggested to be amyloidogenic.^25,26 αSN, with a rigid and compact C-terminus, was found to increase amyloidogenicity.²⁷ These findings suggest that microscopic structural properties in the initial states are important for governing aggregation pathways.

The macroscopic natures of proteins in solution, solubility and supersaturation, have recently been applied to explain protein aggregation.^{11–13,28–31} Proteins dissolved over their solubility form insoluble aggregates in equilibrium with residual monomers. Protein solubility is controlled by solvent polarity using inorganic solvents such as alcohols and also adjusted by salts based on Hofmeister series or temperatures.^12,13,28,32 Supersaturation delays nucleation-dependent phase transition. Supersaturation is a universal definition used to describe the kinetically trapped soluble states of molecules before nucleation when molecules dissolved over their thermodynamic solubility (i.e., threshold concentrations).¹¹ Powerful agitation, such as sonication, is required to efficiently indicate the most thermodynamically stable aggregation type through the promotion of nucleation with the disruption of supersaturated states. Recent studies have also reported that supersaturation is key to understanding protein aggregation in vivo and its role in the pathogenesis of disorders. Supersaturated proteins are common among a number of neurodegenerative diseases,²⁹ and the remodeling of supersaturated proteome during aging predominantly regulates protein aggregation in Caenorhabditis elegans.³³ We previously demonstrated that a phase diagram of protein aggregation was effective for the macroscopic description of protein aggregation, and the combination of alcohol and sonication was markedly useful for examining supersaturation-limited protein aggregation.²⁸

Unstructured (poly)peptides with alcohols transform to α-helical conformations, for instance, corticotropin-releasing factor,³⁴ C-terminus of histones H1 and H5,³⁵ and αSN³⁶ due primarily to the stabilization of intramolecular hydrogen bonds. Alcohols have been widely used for diverse biological membrane-mimicking studies of proteins and peptides, as they provide clear and simple in vitro experimental conditions that biological membrane systems in vivo cannot provide due to technical difficulties.³⁷ In the last two decades, alcohols have shown their superiority to induce amyloidogenesis of proteins such as acylphosphatase (AcP),^38,39 β2m,⁴⁰ and Aβ peptides^41,42 and to trap intermediates for investigations using multiple biophysical techniques. Previous results exhibited that 2,2,2-trifluoroethanol (TFE) is effective for generating Aβ amyloidogenesis.⁴¹ More recently, Aβ peptides have been suggested to self-assemble into amyloid fibrils or nonfibrillar aggregates depending on the concentration of fluorinated alcohols.⁴² However, much remains to be understood about how alcohol-induced structural variations of Aβ monomers impact their complex aggregation. In addition, phase diagram-based solubility and supersaturation, key concepts for the macroscopic understanding of aggregation, have not been systematically introduced to the study of alcohol-induced Aβ aggregation.

Collectively, these findings prompted us to perform an in-depth investigation on the complex protein aggregation process under various conditions based on microscopic and macroscopic viewpoints in order to obtain a more complete understanding of the mechanisms and principles of protein aggregation. Herein, we report our findings regarding diverse structural conversions of Aβ (1–40) (Aβ_1–40) and their relation to aggregation pathways. Fluorinated alcohols and sonication were used to modulate the structure and solubility of Aβ_1–40 monomers and to identify the final molecular species, respectively. We illustrate the importance of helical structures formed at early stages that determine the microscopic Aβ_1–40 aggregation pathway (Figure 1A) and present the mechanisms by which Aβ_1–40 aggregates into the final aggregated species with distinct solvent polarity. We further introduce macroscopic viewpoints, solubility and supersaturation, to understand the monomer structure-dependent amyloid formation (Figure 1A). We confirm that hybrid phase diagrams containing microscopic structural information on precursor proteins and macroscopic phase transition are useful for elucidating the general properties of the aggregation behaviors of folded and unfolded precursor proteins as well as phase separation. Our overall results would also be useful to investigate other amyloid proteins and the aggregation process under diverse conditions. Such studies would be valuable in the development of inhibitors against toxic aggregates and potentially to engineer compounds as potential therapeutics to suppress the pathology of amyloid-related diseases.

Figure 1.

Conformational transition of Aβ_1–40 depending on solvent polarity. (A) Schematic representation of the structural change of random coil-like Aβ_1–40 (left, light blue region) to a largely (left, blue region) and partially (right, light red region) helical fold, which increases and decreases solubility, respectively. Partial helical Aβ_1–40 is in a kinetically soluble state due to supersaturation, and insoluble β-structure-rich amyloid fibrils (right, red region) are thermodynamically formed after the disruption of supersaturation. (B and C) Far-UV CD spectra of Aβ_1–40 at various concentrations of (B) TFE and (C) HFIP. (D) The content of α-helix structures of Aβ_1–40 at each TFE (upper panel) and HFIP (lower panel) concentration was analyzed using the corresponding CD spectra and BeStSel algorithm.⁴⁴ (E and F) Transition diagrams of Aβ_1–40 in (E) TFE and (F) HFIP constructed based on ellipticity values at 222 and 198 nm in B and C. Linear fits of points (E) from 0 to 15% and from 15 to 50% TFE and (F) from 14 to 50% HFIP are represented by dotted lines. (G and H) Structures of Aβ_1–40 in (G) 85% H₂O and 15% TFE as well as (H) 98% H₂O and 2% HFIP after 100 ns simulation are shown. The color of the peptide structures changes from blue to red at the N- and C-termini, respectively. (I–N) AFM and TEM images of the Aβ_1–40 solution at the distinct type and concentration of alcohol were obtained immediately after sample preparation. The scale bars in the AFM and TEM images represent 500 and 100 nm, respectively.

RESULTS AND DISCUSSION

Characterization of Initial Conformational States of Aβ_1–40 in Water/Alcohol Mixtures.

In order to examine initial structural states of Aβ_1–40, a series of far-UV circular dichroism (CD) spectra of Aβ_1–40 were recorded with increases in the concentrations of TFE (Figure 1B) and 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) (Figure 1C). In the absence of alcohol, the Aβ_1–40 solution showed a minimum at ~200 nm, suggestive of the random-coil-like structures as reported by previous studies.^41,43 A narrow distribution of sharp NMR peaks also suggested largely unstructured Aβ_1–40 monomers (Figure S1). Elevations in the concentration of TFE induced changes in far-UV CD spectra with increases in CD signal intensities at shorter wavelengths (<~203, ~208, and ~222 nm, Figure 1B), indicating a structural transition to an α-helix-rich structure and the formation of largely helical structures at 50% TFE. The prediction of the content of secondary structures using the BeStSel algorithm⁴⁴ supported TFE-concentration-dependent increases in the α-helix from ~3 to ~35% (Figure 1D, upper panel). The content of the α-helix (33%) in 40% TFE was well consistent with that of the NMR structure (35%).⁴⁵ Although all spectra at 0–50% TFE appeared to intersect at ~203 nm, they did not share a single isodichroic point. CD spectra at 0–15% and 15–50% showed isodichroic points at ~201 and ~204 nm, respectively. Likewise, the transition diagram constructed by plotting CD intensities at 198 and 222 nm obtained at distinguished TFE concentrations exhibited two linear dependences that intersected at ~15% TFE (Figure 1E), suggesting the existence of different helical structures and partial helical structures at ~15% TFE. Atomic force microscopy (AFM) images confirmed the absence of aggregation from 0 to 50% TFE (Figure 1I–K), suggesting that the structural transition from disordered Aβ_1–40 to different helical conformations did not cause Aβ aggregation at the initial stage.

The change in the CD spectra at various HFIP concentrations was more dynamic (Figure 1C). The AFM of HFIP at low concentrations ranging from 2% to 12% induced the alteration of the random coil-like spectrum to a characteristic spectrum with a single minimum at ~220 nm and the enhancement in the CD intensity at shorter wavelengths, which indicated a β-structure-rich conformational state. Upon further increasing concentrations of HFIP to 14% and 16%, both the negative and positive CD signals were intensified, resulting in the conversion of the overall spectrum to an α-helix-dominated shape. At higher HFIP concentrations of more than 20%, a typical far-UV spectrum of the α-helix-rich structure was obtained. The predicted content of α-helix was as low as ~0.5 to ~4.5% at low HFIP concentrations (0–12%) and was increased up to ~40% at 50% HFIP (Figure 1D, lower panel). All CD spectra showed one isodichroic point at ~200 nm, except for the spectrum without HFIP, suggesting a transition from β to α. The transition diagram displayed a linear dependence between 14% and 50% HFIP (Figure 1F). Transmission electron microscopy (TEM) and AFM images visualized short, curvilinear fibrillar aggregates at 2%, 6%, and 12% HFIP (Figures 1L,M and S2A–G), indicating the rapid formation of β-structured protofibrils. Although differences in CD spectra suggested distinct secondary structures of protofibrils at the different concentration of HFIP, all protofibrils showed similar morphologies (Figures 1C,L,M and S2A–G). High-resolution TEM images revealed that protofibrils at all conditions consisted of a single strand with a width of ~5 nm (Figure S2A–G). At greater than 20% HFIP, no aggregate was detected in TEM and AFM images (Figures 1N and S2H), suggesting Aβ_1–40 with highly helical structures remained soluble at high HFIP concentrations.

Next, we performed a MD simulation to determine the mechanistic and structural arrangement of Aβ_1–40 in mixtures of two different solvents containing 85% H₂O and 15% TFE, where a partially helical intermediate was observed, as well as 98% H₂O and 2% HFIP, where rapid β-rich aggregation occurred at the atomic resolution. Largely unfolded Aβ_1–40 was structurally diverged depending on conditions (Figures 1G,H and S3). In 15% TFE solution, Aβ_1–40 rearranged with the induction of a short helical structure (residues E11–K16) (Figures 1G and S3A). The helical content of 12.5% observed was in good agreement with the CD analysis. Meanwhile, the addition of 2% HFIP also induced a partial helical structure at the residues V24–A30 from largely unfolded structures (Figures 1H and S3B). MD snapshots revealed that the homogeneously distributed TFE/HFIP molecules aggregated surrounding Aβ_1–40 (Figure S4). Therefore, the discerning intermolecular interaction between Aβ_1–40 and organic solvent molecules plays a central role in mediating the initial partial helical conformations of Aβ by reinforcing intramolecular hydrogen bonds.⁴⁶

Observation of Aβ_1–40 Aggregation in Water/Alcohol Mixtures.

The real-time monitoring of Aβ_1–40 aggregation was first performed by the thioflavin T (ThT) assay under quiescent conditions without TFE (Figure 2A). Even after incubation for ~2 days, no change was observed in the ThT fluorescence intensity. The solution of Aβ_1–40 after incubation showed the same far-UV CD spectrum recorded prior to incubation (Figure 2B). These results indicated that amyloid fibrils were not formed. On one hand, sonication rapidly increased the ThT fluorescence intensity, with a lag time of ~1 h and an elongation rate constant of ~10 h⁻¹ (Figure 2L and M), and triggered a far-UV CD spectrum of a typical cross-β structure of amyloid fibrils by showing a minimum at ~220 nm, indicative of the nucleation-dependent amyloid fibrillation of Aβ_1–40 (Figure 2B). The formation of amyloid fibrils was also confirmed by TEM (Figure 2G) and one-dimensional proton NMR spectroscopy based on the disappearance of NMR peaks (Figure S1).

Figure 2.

TFE concentration-dependent amyloid formation of Aβ_1–40 under quiescent and stimulated conditions. (A–I) The aggregation of Aβ_1–40 was traced by (A, C, and E) ThT fluorescence, (B, D, and F) far-UV CD, and (G, H, and I) TEM at various TFE concentrations. The ThT intensities of the Aβ_1–40 solution without agitation (blue triangle) and with sonication (red circle) were plotted as a function of time. Solid lines represent the fit curves of the kinetics of Aβ_1–40 amyloid fibrillation. Dotted lines were drawn as an eye-guide only. Inserts in A and C are the magnified profiles of rapid kinetics of Aβ_1–40 fibrillation with sonication. Far-UV CD spectra of Aβ_1–40 were recorded after incubation without agitation (blue line) and with sonication (red line). The CD spectra of Aβ_1–40 measured soon after sample preparation are shown as a comparison (black line). TEM images were taken from Aβ_1–40 samples after incubation with sonication. The concentration of TFE used was displayed above the TEM images. Scale bar = 400 nm. (J–M) The lag time (J and L) and elongation rate constant (K and M) of Aβ_1–40 fibrillation in TFE/water mixtures without agitation (J and K) and with sonication (L and M) are displayed. “n.d.” denotes the TFE concentrations at which no enhancement in the ThT fluorescence intensity was observed. Error bars indicate the standard deviation of three independent measurements.

An increase in the ThT fluorescence intensity and the reduced lag time were broadly observed with the far-UV CD spectra of cross-β amyloid architectures when the concentration of TFE was elevated from 0 to 20% under quiescent conditions (Figures 2C,D and S5A–C), indicating the acceleration of amyloid formation. Further enhancement in the TFE concentration to more than 25% caused no change in the ThT-based kinetics and CD spectra, which suggested that amyloid generation did not occur (Figures 2E,F and S5D,E). Thus, our results presented the minimum lag time and maximum elongation rate at ~15% TFE (Figure 2J and K).

Aβ_1–40 in the 5–25% TFE solution, subjected to sonication, showed a noticeable reduction in the lag times and a steeper elongation phase with the far-UV CD spectra of the cross-β structures of amyloid fibrils as well as fibrillar aggregates in the TEM image (Figures 2C,D,H and S5A–D), supporting the acceleration of amyloid formation from that under alcohol-free conditions. Although the effects of sonication lowered the difference between lag times or elongation rates, the TFE-concentration dependence of two kinetic parameters was similar to that under quiescent conditions with indication of minimum lag times at 0–20% TFE and a maximum elongation rate at 15% TFE (Figure 2L and M). In addition, amyloid fibrils formed both with and without 15% TFE decreased the viability of human neuroblastoma cells by ~20%, suggesting the cytotoxicity of amyloid fibrils (Figure S6). At greater than 40% TFE, sonication did not affect the ThT fluorescence (Figure 2E).

It is commonly accepted that ThT binds to surface side-chain grooves of amyloid fibrils largely through hydrophobic interactions.⁴⁷ Thus, the increased hydrophobicity at higher TFE concentrations may disrupt the interaction between ThT and Aβ_1–40 amyloid fibrils, although large enhancements of the ThT fluorescence intensity have been detected when fibrils of lysozyme,¹² insulin,¹³ and AcP³⁸ formed in solutions containing more than 40% TFE. Far-UV CD results showed that Aβ_1–40 monomers remained largely unstructured throughout the incubation, which was in good agreement with the absence of aggregates in the TEM micrograph (Figures 2F,I and S5E). These results supported that Aβ_1–40 amyloid fibrillation was blocked at the concentrations of TFE larger than 40%.

The addition of HFIP led to a characteristic aggregation profile of Aβ_1–40 (Figure 3). The ThT fluorescence at 2% HFIP, recorded immediately after the initiation of the ThT assay under quiescent conditions, showed the high intensity, proposing the formation of ThT-positive aggregates (Figure 3A). Moreover, there was no increase in the ThT fluorescence intensities of a 2% HFIP solution without Aβ_1–40 (Figure S7), further suggesting that the initial high ThT intensity in Figure 3A was attributed to Aβ_1–40 aggregation. Based on short rod-like curvilinear fibrillar morphologies (Figure 1L) and the far-UV CD spectrum of a β-structure, which exhibited a minimum at ~218 nm (Figure 3B), β-structured protofibrils were generated. A sigmoidal increase in the ThT fluorescence was observed after a lag time of ~6 h. After incubation, long and thick mature amyloid fibrils were shown in the AFM image (Figure S8A) with the far-UV CD spectrum of a cross-β structure (Figure 3B). Strong agitation with sonication significantly accelerated the kinetic process of the transition from protofibrils to mature amyloid fibrils, resulting in its completion within ~2 h (Figure 3A). The sonicated samples displayed the signature of a β-structure molecular species with a minimum at ~215 to ~220 nm (Figure 3B). Both TEM (Figure 3C) and AFM images (Figure S8B) presented that fibrillar aggregates were produced and sonication clearly generated short aggregates. Taken together, these results indicated the formation of amyloid fibrils. The conversion rate constants from proto- to mature fibrils were 0.15 ± 0.05 and 2.49 ± 0.70 h⁻¹ in the absence and presence of sonication, respectively.

Figure 3.

HFIP concentration-dependent amyloid formation of Aβ_1–40 under quiescent and stimulated conditions. (A–I) The aggregation of Aβ_1–40 was traced by (A, D, and G) ThT fluorescence, (B, E, and H) far-UV CD, and (C, F, and I) TEM at various HFIP concentrations. The ThT intensities of the Aβ_1–40 solution without agitation (blue triangle) and with sonication (red circle) were plotted as a function of time. Solid lines represent the fit curves of the conversion of protofibrils to amyloid fibrils. Dotted lines were drawn as an eye-guide only. Inserts in A and D are the magnified profiles of rapid kinetics of Aβ_1–40 fibrillation with sonication. Far-UV CD spectra of Aβ_1–40 were recorded after incubation without agitation (blue line) and with sonication (red line). The CD spectra of Aβ_1–40 measured after sample preparation are shown for a comparison (black line). TEM images were taken from the Aβ_1–40 samples after incubation with sonication. Scale bar = 100 nm.

At 6% HFIP without sonication, the initial high ThT fluorescence intensity (Figure 3D) with rod-like curvilinear fibrillar aggregates (Figure 1M) and the far-UV CD spectrum of a β-structured pattern (Figure 3E) are shown, supporting the presence of ThT-positive β-structured protofibrils. In addition, similar to the result at 2% HFIP, the sample solution containing 6% HFIP without Aβ_1–40 did not impact the initial ThT fluorescence (Figure S7). The intensity of ThT fluorescence did not change even after ~35 h. The far-UV CD spectrum of a β-structured pattern was very similar regardless of incubation (Figure 3E), and the AFM image exhibited small aggregates (Figure S8C), which suggested that protofibrils still remained. However, sonication caused a rapid and marked increase in the ThT intensity and completed the reaction at ~2 h with the conversion rate constant of 2.21 ± 0.85 h⁻¹ (Figure 3D). The CD spectrum of the sonicated sample solution showed a β-structured pattern with an increased positive CD signal at a low wavelength (<210 nm) (Figure 3E). TEM (Figure 3F) and AFM images (Figure S8D) at the ThT end point states presented short amyloid fibrils due to sonication.

Although strong ultrasonic effects concealed the difference in aggregation kinetics, faster amyloid fibrillation at 2% HFIP than that at 6% HFIP under quiescent conditions suggested that the colloidal and kinetic stabilities of protofibrils at 2% HFIP were lower than those at 6% HFIP. At a high HFIP concentration of 50%, the ThT fluorescence without sonication maintained a very low intensity during the incubation (Figure 3G). In addition, an α-helical far-UV CD spectrum did not change its pattern (Figure 3H). Any aggregates were not found in the TEM image (Figure 3I).

Characterization of Aβ_1–40 Oligomers and Their Conversion to Amyloid Fibrils.

In order to determine various types and the interconversion of aggregates, we prepared an additional type of aggregates, Aβ_1–40 oligomers, based on previous studies (Figure 4).^48,49 Aβ_1–40, which was freshly dissolved in 100% HFIP, was diluted to 10% HFIP using deionized water (see details in Methods). The far-UV CD measurement of the sample solution showed a β-structure-representing spectrum, which was similar to that of protofibrils, but distinguishable from that of mature amyloid fibrils (Figure 4A). The ThT fluorescence intensity of aggregates was noticeably lower than those of protofibrils and amyloid fibrils (Figure 4B). The TEM and AFM images indicated the production of spherical aggregates (Figure 4C), which were also morphologically different from fibrillar aggregates, protofibrils, and amyloid fibrils (Figures 1–3 and S2 and S8). High-resolution TEM images identified diverse types of oligomers with different morphologies and sizes (Figure S9) that would not be detected using AFM. For example, a donut-shaped oligomer that was supposed to contain β-barrel-like structures⁵⁰ was observed (Figure S9C). Oligomers with a spheroid morphology (Figure S9G) or irregular bulges on the surface (Figure S9H) were also detected, consistent with previous findings that different types of oligomers were detected both in vitro and in vivo.^51,52 The ELISA using an oligomer-specific antibody (A11)⁴⁹ unambiguously confirmed the formation of Aβ_1–40 oligomers showing a high intensity for the antibody reaction at 450 nm (Figure 4D). Taken together, these results indicated the oligomerization of Aβ_1–40, which was rationalized by the following two reasons: (1) solubility of Aβ_1–40 monomers in 100% HFIP decreased due to the dilution to 10% HFIP, which triggered aggregation such as oligomerization; (2) intermolecular hydrogen bonding for β-structured Aβ_1–40 oligomers at 10% HFIP became more thermodynamically favorable than intramolecular hydrogen bonding for helical Aβ_1–40.

Figure 4.

Characterization of Aβ_1–40 oligomers. (A) Far-UV CD spectra of Aβ_1–40 monomers (black line), oligomers (magenta line), protofibrils (green line), and mature amyloid fibrils (red line). (B) Fluorescence intensities of ThT at 485 nm for Aβ_1–40 monomers, oligomers, protofibrils, mature amyloid fibrils, and buffer alone. (C) TEM (upper panel) and AFM (lower panel) images of Aβ_1–40 oligomers. The scale bars of TEM and AFM represent 200 and 500 nm, respectively. (D) Detection of Aβ_1–40 oligomers using the sandwich ELISA based on absorbance at 450 nm. Absorbance of monomers was recorded for the control. (E) Distribution of the hydrodynamic radii of Aβ_1–40 oligomers, obtained by dynamic light scattering measurements. (F) Sedimentation velocity profiles of Aβ_1–40 oligomers optically detected at 220 nm. Fitted sedimentation boundaries measured at an interval of 15 min are shown. (G) Distribution of the sedimentation coefficient of Aβ_1–40 oligomers derived from the sedimentation velocity data in F.

The sizes of Aβ_1–40 oligomers were then analyzed (Figure 4E–G). Dynamic light scattering measurements exhibited a narrow distribution of the hydrodynamic radius of Aβ_1–40 oligomers with a diameter of ~10 to ~50 nm (Figure 4E), which was consistent with the sizes of oligomers observed in the TEM and AFM images. High molecular weights of oligomers induced NMR peak broadening (Figure S1). A clear sedimentation profile indicated the absence of large Aβ_1–40 aggregates (Figure 4F). Sedimentation coefficient (s) values obtained by the sedimentation velocity measurement of analytical ultracentrifugation were distributed from 0 to ~40 S with indication of one small and two large peaks (Figure 4G). Two s-values, ~5.2 and ~24.7 S, were representatively selected to assess the molecular weights and number of Aβ_1–40 monomers consisting of oligomers. The molecular weights of Aβ_1–40 oligomers were calculated to be ~70 and ~740 kDa. Therefore, ~17 and ~170 Aβ_1–40 peptides formed small and large oligomers, respectively.

The solutions of Aβ_1–40 oligomers were incubated for ~2 days without agitation, and no changes were observed in the ThT fluorescence intensity (Figure 5A). We also did not detect any significant differences in the far-UV CD spectra that depended on incubation (Figure 5B). Extensive sonication was introduced; however, no significant alteration was observed in ThT fluorescence over incubation time despite a small initial decrease (Figure 5A). Additionally, a similar pattern for the CD spectrum with the reduced intensity was obtained (Figure 5B). All TEM images exhibited Aβ_1–40 oligomers with spherical morphologies (Figure 5C–E). Clustered oligomers that may decrease the CD intensity were shown (Figure 5E), and this may be due to sonication.

Figure 5.

Context-dependent aggregation of Aβ_1–40 oligomers. (A–K) The aggregation behaviors of Aβ_1–40 oligomers were observed by (A and F) ThT fluorescence, (B and G) far-UV CD spectra, (C–E, H and I, and K) TEM, and (J) AFM in the (A–E) absence and (F–K) presence of 10 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl. The ThT intensities of Aβ_1–40 oligomers without agitation (blue triangle) and with sonication (red circle) were plotted as a function of time. Dotted lines were drawn as an eye-guide. The inset in F represents a magnified kinetic trace of the rapid increase in the ThT fluorescence intensity with sonication. The far-UV CD spectra of Aβ_1–40 oligomers were recorded after incubation without agitation (blue line) and with sonication (red line). The CD spectra of Aβ_1–40 oligomers measured before incubation (black line) are displayed for a comparison. The scale bars in C–E and H–K indicate 200 and 100 nm, respectively.

The context-dependent transformation of Aβ_1–40 oligomers was then examined (Figure 5F–K). We added 10 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl to the solution of oligomers and observed the time-course of variation of the ThT fluorescence (Figure 5F). The ThT fluorescence intensity at time zero was more than 8-fold larger than that of oligomers before changing the condition, which indicated that the change to a physiological condition rapidly increased the ThT fluorescence intensity of the oligomer solution within the experimental dead time (~5 min) (Figure 5A and F). In contrast, the addition of 10 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl to solution without Aβ_1–40 oligomers did not increase the ThT fluorescence intensity, which ruled out the possibility that a solvent effect caused the enhancement in the ThT fluorescence value (Figure S10). The pattern of the far-UV CD spectrum was similar to that of oligomers with a decrease in the CD intensity (Figure 5G). The TEM and AFM images exhibited short curvilinear fibrillar aggregates (Figure 5H and J), which indicated the burst formation of Aβ_1–40 protofibrils, i.e., the conformational conversion from oligomers to protofibrils.

Incubation under quiescent conditions presented a gradual increase in the ThT fluorescence intensity with saturation at ~35 h (Figure 5F), and the final product displayed the CD spectrum of β-rich structures (Figure 5G). The TEM image clearly showed the production of mature Aβ_1–40 amyloid fibrils (Figure 5I). Sonication strongly promoted the reaction within ~2 h with indication of the increased fluorescence intensity of ThT and the CD signal (Figure 5F and G). Initial decreases in the ThT fluorescence intensity were observed in the presence and absence of sonication (Figure 5F). Amyloid fibrils were detected in the TEM image of the samples after incubation with sonication (Figure 5K). The conversion rate constants of prototo mature fibrils were 0.06 ± 0.02 and 2.26 ± 0.50 h⁻¹ without and with sonication, respectively, which were similar to those in water/HFIP mixtures.

In order to examine effects of environmental factors on the conversion of oligomers to other types of aggregates in more detail, the aggregation process of Aβ_1–40 oligomers, formed at pH ~4, with and without 10 mM sodium phosphate buffer (pH 7.5) at the various concentration of NaCl (0–2000 mM) was investigated (Figure 6). The addition of 10 mM sodium phosphate buffer (pH 7.5) without NaCl induced a rapid increase in the ThT fluorescence intensity at the initial stage (Figure 6A), which indicated the formation of protofibrils as observed in Figure 5. Incubation with continuous shaking led to the gradual increase in the ThT intensity with saturation at ~20 h. The far-UV CD spectrum showed a typical pattern of β-structure-rich conformations (Figure 6B), and the TEM image exhibited the formation of mature amyloid fibrils (Figure 6G), indicating that the pH jump from ~4 to 7.5 is enough to cause the conversion from oligomers to mature amyloids via protofibrils.

Figure 6.

Effects of the pH and NaCl concentration on the transition of Aβ_1–40 oligomers. (A–M) Aggregation behaviors of Aβ_1–40 oligomers at the various concentrations of NaCl were observed using (A, C, D, and F) ThT fluorescence, (B and E) far-UV CD spectra, and (G–M) TEM in the (A–C and G–K) presence and (D–F, L, and M) absence of 10 mM sodium phosphate buffer (pH 7.5). Solid lines in A represent the fit curves for the kinetic analysis of the conversion to amyloid fibrils. The rate constant for the conversion is shown in C and F. Error bars represent the standard deviation of three independent measurements. “n.d.” denotes the NaCl concentrations where the protofibril-to-amyloid fibril conversion did not occur. The far-UV CD spectra of Aβ_1–40 oligomers were recorded after the 24 h incubation. The concentrations of NaCl used are displayed in A, B, and G–M. The scale bars in G–M indicate 200 nm.

At the concentration of NaCl ranging from 50 to 250 mM, the initial ThT fluorescence was similar to that in the absence of NaCl, implicating the formation of protofibrils, and the gradual enhancement in the ThT fluorescence intensity was observed. CD spectra showed the formation of β-structure-rich aggregates (Figure 6B), and TEM images exhibited long fibrillar aggregates (Figure 6H and I), indicating that the final product is amyloid fibrils. The addition of the high concentrations of NaCl (1000 and 2000 mM) resulted in no increase in the ThT fluorescence even after ~24 h (Figure 6A), and large aggregates of protofibrils that may decrease the CD intensity (Figure 6B) were predominant without amyloid fibrils in the TEM images (Figure 6J and K). The kinetic analyses revealed an optimum concentration of NaCl, i.e., ~100 mM, to convert oligomers to amyloid fibrils through protofibrils (Figure 6C). We speculated that the initial decrease in the ThT fluorescence was related to a conversion of protofibrils to form amyloid fibrils.

On the other hand, no increase in the ThT intensity was observed throughout the incubation with the addition of NaCl (50–2000 mM NaCl) alone (Figure 6D) by showing β-structured CD spectra (Figure 6E). TEM images visualized large aggregates without any morphology of protofibrils and amyloid fibrils at both 100 and 2000 mM NaCl (Figure 6L and M). All these results indicated that the addition of NaCl to oligomer samples without a pH shift induced the formation of large amorphous aggregates, which may consist of small β-structured oligomers instead of amyloidogenic aggregation via protofibrillation.

Amyloid Plaque Deposition in Vivo by the Injection of Aβ_1–40 Oligomers and Protofibrils.

In order to investigate whether Aβ_1–40 oligomers or protofibrils formed in vitro are involved in Aβ amyloid deposition, i.e., amyloid plaque formation, in vivo, we bilaterally injected these two types of Aβ_1–40 aggregates and vehicle controls into the hippocampus CA1 regions of wild-type adult mice (Figure 7A). At 24 h after intrahippocampal injections, the mice were perfused and fixed, and immunohistochemistry was performed with anti-4G8 antibodies, which have been used to visualize Aβ plaques in vivo (Figure 7B).⁵³ Interestingly, we found that oligomer- or protofibril-injected mice significantly increased 4G8-positive immunoreactivity in the hippocampus CA1 region compared with vehicle-injected mice (Figure 7B), implicating the formation of amyloid plaques. The quantitative analysis revealed that 4G8 fluorescence intensities of mice with the injections of oligomers and protofibrils were ~10 fold higher than those with vehicle injections (Figure 7C). The 4G8 fluorescence intensity of protofibril-injected mice was higher than that of oligomer-injected mice. These results may suggest that oligomers and protofibrils formed in vitro can be changed to amyloid fibrils for Aβ plaque deposition in vivo. The lower fluorescence intensity of oligomer-injected mice than that of protofibril-injected mice suggested that protofibrils might convert to amyloid fibrils more easily than oligomers in vivo, as oligomers undergo the conversion to amyloid fibrils via protofibrils. We speculate that complicated environments in brains may impair the oligomer-to-protofibril transition to some extent.

Figure 7.

Immunohistochemical examination of the deposition of amyloid plaques in the hippocampal CA1 region. (A) Illustration of the cross section of the mouse brain to show the injection site (CA1). (B) Immunohistochemical results at 24 h after the injection of vehicles (top row), Aβ_1–40 oligomers (middle row), and protofibrils (bottom row) to the bilateral hippocampal CA1 regions. DAPI (left column), 4G8 immunostaining (middle column), and magnification of CA1 regions (right column) are shown. (C) Quantification of 4G8-positive amyloid plaques in the hippocampal CA1 region (n = 4 mice per group). Data represent mean ± SEM (***p < 0.001).

Microscopic Characterization of Aβ_1–40 Aggregation Pathways.

Amyloid formation has been extensively studied for many different amyloid proteins and peptides; however, the different mechanisms and pathways underlying the aggregation process have not been well established. Recent studies have demonstrated the use of variables and conditions, such as temperature,⁵⁴ lipid bilayer/nanodiscs,^55,56 and chemical molecules^57–59 such as phthalocyanine tetrasulfonate, 2,8-bis(2,4-dihydroxyphenyl)-7-hydroxyphenoxazin-3-one, and heparin sulfate to probe the aggregation pathways. In our present studies, we aimed to comprehensively reveal the aggregation mechanisms and pathways employing Aβ_1–40 as a model system and a variety of biophysical experiments and MD simulation.

Six possible microscopic pathways are revealed for Aβ_1–40 aggregation (pathways 1 to 5) at the atomic and molecular level (Figures 8 and S11) (Table 1). In aqueous media, the formation of amyloid fibrils (β_AF) from random-coil conformations of Aβ_1–40 (RC) was slow under quiescent conditions due to a high energy barrier, and sonication accelerated cytotoxic fibrillation (Figure S11) via a decrease in this energy barrier (pathway 1, RC-β_AF transition). At moderate concentrations of TFE, partial helical structures (α^P) with relatively high stability accumulated as a kinetic intermediate that proceeded to β-structured amyloid fibrillation (pathway 2, RC-α^P-β_AF transition) with acceleration by sonication. Aβ_1–40 with a helical content of ~15 to ~20% at 15–20% TFE may be optimal for toxic amyloid formation, which is consistent with previous findings.⁴¹ The MD simulation in 15% TFE revealed a partial helical structure of Aβ_1–40 composed of E11, V12, H13, H14, Q15, and K16, which may be responsible for the productive nucleation for amyloidogenesis through helix–helix interactions.⁶⁰ Notably, the introduction of proline, known as a helix breaker, at residue 14 significantly retarded the lag time for nucleation.⁶¹ The extension of this partial helix from H13 to D23 induced the generation of high molecular weight assemblies.⁶² The aggregation of α-helical structures in order to proceed to amyloid formation was observed in polyQ fibrillation.⁶³ Aggregation-prone partial helical structures of αSN,^31,36 hIAPP,⁶⁴ a fragment of β2m,⁶⁵ and apoA-I⁶⁶ have also been suggested.^60,67

Figure 8.

Scheme of the microscopic aggregation pathway of Aβ_1–40. Various aggregation pathways are shown and guided by arrowed curves in distinct colors: pathway 1 (gray), pathway 2 (black), pathway 3 (green), pathway 4 (blue), pathway 4′ (red), and pathway 5 (magenta). α^P in pathways 2 and 3 obtained from MD simulations was used. α^P in pathways 4 and 4′ and α^H in pathways 4, 4′, and 5 were conceptually drawn. α^P in pathways 3, 4, and 4′ is too transient to be detected using general experimental tools used here. The markedly rapid reactions of which kinetic parameters could not be obtained are represented using bold arrowed curves. The reactions accelerated by sonication are highlighted with cartoons. Kinetic parameters (i.e., the rate constant of fibril growth and conversion) in the absence and presence (values in parentheses) of sonication are shown. “n.d.” denotes the reaction that did not show an increase in the ThT fluorescence during incubation time used here.

Table 1.

Summary of Various aβ_1–40 Aggregation Pathways

pathway	structural transition and aggregation scheme
pathway 1	random coil (RC) → β-structured amyloid fibril (βAF) (RC-βAF transition)
pathway 2	random coil → partial α-helical structure (αP) → β-structured amyloid fibril
	(RC-αP-βAF transition)
pathway 3	random coil → partial α-helical structure → β-structured protofibril (βPF) → β-structured amyloid fibril
	(RC-αP-βPF-βAF transition)
pathway 4	random coil → highly α-helical structure (αH) → partial α-helical structure → β-structured oligomer (βOG) → β-structured large aggregate (βLA) (RC-αH-αP-βOG-βLA transition)
pathway 4′	random coil → highly α-helical structure → partial α-helical structure → β-structured oligomer → β-structured protofibril → β-structured amyloid fibril (RC-αH-αP-βOG-βPF-βAF transition)
	αH and αP indicate highly and partially α-helical structures at 100% and 10% HFIP, respectively; italics indicate molecular species at a physiological pH value (e.g., pH 7.5) with NaCl in the range 0–250 mM
pathway 5	random coil → highly α-helical structure (RC-αH transition)
	αH represents highly α-helical structures at high alcohol concentrations.

In pathway 3 (RC-α^P-β_PF-β_AF transition), the addition of small amounts of HFIP (2%) triggered a dynamic structural transition from RC. α^P, which was undetectable in a general experimental time scale, must be unstable (i.e., transient) and aggregation-prone with a shallow energy minimum. α^P showed one short helical region (V24, G25, S26, N27, K28, G29, and A30) (Figure 1H, right panel) which corresponds to the turn region (V24–A30) between two β-strands (V12–V24 and A30–V40) in the hydrophobic core of amyloid fibrils.⁶⁸ Destabilization of turn structures of Aβ monomers (V24–K28) induced largely disordered and helical structures which promoted oligomerization or protofibrillation,^69,70 implicating the importance of short helical region (V24–K28) in determining an aggregation pathway. The enhanced hydrophobicity of the partial helical structure in 2% HFIP, relative to 15% TFE, might strengthen the helix–helix interaction, causing rapid aggregation. ThT-positive β-structured protofibrils (β_PF) are apparent kinetic intermediates of misfolded aggregates. No similar structure of α^P in 2% HFIP is available up to now due probably to this significant aggregation propensity. Cytotoxic amyloid fibrils (Figure S6) emerged as a final molecular species with the disappearance of β_PF (Figure 3), and the transition was again promoted by sonication. Since similar rapid formation of β-structured aggregates and ThT kinetic profiles were observed at 6% and 12% HFIP (Figures 1 and 3), we speculate that sticky helical conformations may also transiently exist at 6–12% HFIP and follow pathway 3.

In pathway 4 (RC-α^H-α^P-β_OG-β_LA transition) and pathway 4′ (RC-α^H-α^P-β_OG-β_PF-β_AF transition), the RC underwent a transition to ThT-negative β-structured oligomers (β_OG) via α^P at 10% HFIP, which might be distinguished from a highly α-helical structure at 100% HFIP (α^H) (Figure S12). In the absence of pH jump, further incubation with sonication or modulation of ionic strength led to the production of nonfibrillar large aggregates accommodating β structures (β_LA). A pH jump from ~4.0 to 7.5 with moderate concentrations of NaCl changed the aggregation mechanism from pathway 4 to pathway 4′. β_OG transformed rapidly to ThT-positive β_PF, which eventually formed β_AF. High concentrations of NaCl with a pH jump blocked pathway 4′, while the absence of a pH jump kept pathway 4.

It has been long debated whether oligomers represent on-pathway intermediates to amyloid fibril formation or off-pathway products.^71,72 Aβ peptides have shown the formation of both on- and off-pathway oligomers depending on environmental conditions such as metal ions⁷³ and small molecules⁵⁹ as well as preparation methods.^49,74 Herein, β_OG was unable to convert to amyloid fibrils even with extensive treatment of sonication, which indicated that β_OG is an off-pathway aggregate, as it represents a deep minimum in the energy landscape (Figure S11). Previous studies suggested that a significant structural conversion was required for the transition from oligomers to amyloid fibrils,⁷⁵ and this transition was determined by the shape, compactness, and amount of β-structures in oligomers.⁷⁶ Increases in stability of β_OG due to the formation of β structures might decrease an energy level (i.e., thermodynamic stabilization) by creating a high energy barrier (i.e., kinetic stabilization) for the structural reorganization of oligomers toward amyloid aggregation. However, the shift to physiological conditions led to a pathway toward amyloid formation of β_OG via protofibrillation in vitro (Figure 6) and possibly in vivo (Figure 7).

pH has been considered to play a crucial role in determining the aggregation pathways of proteins. It has been observed that pH jumps triggered amyloid aggregation of hIAPP,⁷⁷ human procarboxypeptidase A2,⁷⁸ and the SH3 domain.⁷⁹ We herein demonstrated that the pH jump to 7.5 was a prerequisite for the transition from the end product of Aβ_1–40 oligomers to mature amyloid fibrils via intermediates of protofibrils. On the other hand, studies of the salt effects on protein aggregation have been extensively performed.^80–82 Changes in β2m aggregation pathways from amyloid formation to amorphous aggregation due to the increase in the concentration of NaCl have been observed.¹¹ These observations suggested that neutral salts such as NaCl may be important for determining the aggregation pathway. In this study, we showed that a physiological condition (neutral pH, ionic strength of 125 mM, and 37 °C) was critical for amyloid formation of Aβ_1–40. The injection of Aβ_1–40 oligomers and protofibrils to brains of mice might cause the formation of amyloid plaques (Figure 7).

Pathways 3 and 4′ demonstrated that Aβ_1–40 amyloid formation was in equilibrium because the final product of both was amyloid fibrils irrespective of the distinct aggregation pathway. On the other hand, pathway 5 (RC-α^H transition) prevailed at high alcohol concentrations (>40% TFE and >30% HFIP). Predominantly α-helical structures (α^H) were highly stable and did not form any aggregates. Highly helical structures of Aβ peptides in the high concentration of alcohol (e.g., 40% TFE⁴⁵ or 80% HFIP⁸³) or the presence of sodium dodecyl sulfate (SDS)⁸⁴ showed no aggregation, thereby providing sufficient time for structural determination using NMR spectroscopy not like α^P in pathway 3.

Aβ_1–40 exhibited a similar pathway alteration in a manner that depended on the concentration of SDS.⁸⁵ In the absence of SDS, Aβ_1–40 was fibrillated with pathway 1. An increase in the concentration of SDS resulted in pathway 4′ via pathway 3. Similar findings of multiple pathways were also reported for the aggregation of acid-induced disordered β2m and intrinsically disordered αSN.^36,40,86,87 Unstructured β2m indicated pathway 1 without TFE and pathway 2 with an increase in the concentration of TFE. αSN was previously shown to self-assemble with variations in the concentrations of SDS, TFE, or lipids in model membranes.^{31,36,67,86,87} αSN presented the following pathway transition with an increase in the amount of SDS or lipids: pathway 1 → pathway 3 → pathway 5.

A recent study reported that the distinct structural states of familial mutants of αSN were responsible for different aggregation pathways,⁸⁸ which suggested the importance of initial conformations for aggregation. Thus, we propose the importance of α-helical structures at the early stages of aggregation for controlling Aβ_1–40 aggregation pathways. Amyloid-competent partial helical structures predominated at moderate and low concentrations of TFE and HFIP, respectively. These helices may be less prominent in aqueous media, which results in slow fibrillation. It is worth noting that the position of partial helical structures in Aβ peptides is also important for the aggregation pathway and amyloidogenicity as suggested in pathways 2 (α^P in 15% TFE) and 3 (α^P in 2% HFIP). Highly helical structures that are markedly less active may be present at high alcohol concentrations, resulting in the absence of amyloidogenesis.

It is worth noting that fluorinated alcohols also have the capability to stabilize the intermolecular β-sheet structures.^39,89 Enhancements in intermolecular hydrogen bonds due to the stabilization of β-sheet structures have been suggested to be a driving force for the promotion of AcP self-assembly at TFE concentrations ranging from 5% to 25%³⁹ and αSN amyloid generation in 2.5–10% TFE.⁸⁹ Thus, it was concluded that fluorinated alcohols stabilize first aggregation-prone partial helical structures with the enhancement of intramolecular hydrogen bonds in terms of kinetics, however, thermodynamically stabilize the final end product of amyloid fibrils with β-sheet structures due to enhanced intermolecular hydrogen bonds as shown in pathway 2.

Sonication has been shown to produce cavitation bubbles that provide air–water interfaces.¹¹ The helical folding of amphiphilic peptides and the deformation of amyloid fibrils have been observed at the air–water interface.^90,91 Therefore, the fibrillation-promoting effects of sonication may be dually attributed to the accumulation and condensation of active helical conformations for nucleation (pathways 1 and 2) and to the remodeling of intermediate aggregates to amyloid fibrils (pathways 3 and 4′).

Fast helix folding has been observed in short peptides and globular proteins for the formation of intramolecular hydrogen bonds.⁹² An α–β transition during protein folding has been demonstrated in a limited number of cases, such as β-lactoglobulin⁹³ and the intestinal fatty acid-binding protein.⁹⁴ Random coil-α–β transition, however, may be a generic property of intermolecular protein aggregation when forces to form intramolecular hydrogen bonds are strong, for example, under conditions of low solvent polarity or strong electrostatic interactions such as membrane binding. Our microscopic mechanical models may be useful for understanding Aβ_1–40 aggregation mechanisms on membranes in nonpolar environments. Fast folding to partial helical states coupled to rapid aggregation may be beneficial for Aβ_1–40 in order to protect it from proteolysis; however, this impairs proteostasis.

A large number of studies on structural details and toxic natures of protein aggregates have been reported in the past decade with advances in methods and techniques.^6–9,95,96 However, the fundamental question on whether aggregates generated in vitro display structural and/or toxic features similar to those accumulated in patients is still open to discussion. Recent structural studies revealed that heparin-induced tau filaments had a larger core region⁹⁷ than those obtained from patients of Alzheimer’s and Pick’s diseases.^6,95,96 Our in vivo results implied that the injection of either Aβ oligomers or protofibrils generated in in vitro systems induced the formation of amyloid plaque in brains of mice, although the relevance of in vitro aggregates to aggregates formed in patients of Alzheimer’s disease still needs to be elucidated. Aβ_1–40 aggregation and resulting plaque formation in vivo are much more complicated processes than those in our in vitro system due to the complex intra- and extracellular environments. Thus, future investigation on complicated Aβ_1–40 aggregation pathways in vivo with the advances in biological or physiological techniques will help us clarify the mechanism underlying Aβ_1–40 self-assembly.

Phase Diagrams of Aβ_1–40 Aggregation.

We previously demonstrated that phase diagrams were very useful for comprehensively illustrating the protein aggregation of lysozyme,¹² insulin,¹³ and β2m³⁰ in a macroscopic manner and further showed a potential link between macroscopic views of phase equilibrium and microscopic views of structural states using phase diagrams of cytochrome c and αSN aggregation.^28,31,67 Through our current studies, we expanded the phase diagram-based general description of protein aggregation in terms of both viewpoints.

The sonication of water/TFE mixtures broadened the amyloid-forming region of Aβ_1–40 (0–25% TFE) (Figure 9A) over that under quiescent conditions (5–20%) (Figure 9B), indicating the high metastability of supersaturation at 0 to ~5% and ~20 to ~25% TFE. Aβ_1–40 slowly formed amyloid fibrils at 2% HFIP and protofibrils at 6–12% HFIP without agitation (Figure 9C). Sonication rapidly transformed protofibrils to mature amyloid fibrils (Figure 9D). These results suggested the high metastability of supersaturation at 0 to ~2% HFIP and possibly the absence of apparent supersaturation at ~6% HFIP. Low colloidal stability of protofibrils caused transformation to thermodynamically stable amyloid fibrils. At high concentrations of alcohol (>40% TFE and >30% HFIP), Aβ_1–40 remained soluble regardless of extensive sonication. These phase behaviors of Aβ_1–40 may be macroscopically explained by the metastability of supersaturation with solubility.²⁸ The metastability of supersaturation may be similar to an energy barrier for protein aggregation such as amyloid formation (Figure S11). Thus, strong supersaturation prolongs amyloid formation due to a high energy barrier for productive nucleation. Sonication induces effectively the collapse of supersaturated states, revealing amyloidogenicity.

Figure 9.

Macroscopic phase diagrams of Aβ_1–40 aggregation in water/alcohol mixtures. (A–D) Phase diagrams of Aβ_1–40 in the (A and C) absence and (B and D) presence of sonication. Cartoons of helical monomers, amyloid fibrils, and protofibrils are illustrated. Colors and symbols represent molecular species: (blue region and ▲) soluble monomers; (red region and ■) mature amyloid fibrils; (yellow region and ◆) protofibrils. Each symbol indicates the concentration of Aβ_1–40 and alcohol used. The α-helical contents of the initial states of Aβ_1–40 before aggregation, analyzed by the BeStSel algorithm,⁴⁴ are displayed at the top of the phase diagrams. The dashed and dash-dotted lines at each phase diagram represent conceptual solubility and boundary curves, respectively. Aβ_1–40 aggregation in the presence of 100 mM NaCl at pH 7.5 with (A and B) TFE and (C and D) HFIP.

Higher solubility greater than 25% TFE protects Aβ_1–40 from aggregation and is not under the control of supersaturation and agitation. Soluble helical folds at high alcohol concentrations may be stabilized by enthalpic gains from the formation of intramolecular hydrogen bonds. Consequently, the low aggregation propensity of largely helical structures will also be ascribed to high solubility as a macroscopic thermodynamic property or to an amyloid-protected helical conformation as a microscopic structural feature.

The incorporation of the initial helical percentage in phase diagrams provided insights into the aggregation behaviors of unfolded and structured proteins (Figures 9 and S13). Phase diagrams with sonication showed that Aβ_1–40 with 3–31% initial helical structures at 0–25% TFE (Figure 9B) and acid-denatured β2m with 0–38% helical structures at 0–50% TFE aggregated to amyloid fibrils (Figure S13A and B).⁴⁰ As shown by phase diagrams of the aggregation of native globular proteins, unfolded structures with helix contents of ~20 to ~50% for lysozyme¹² (Figure S13C and D) and ~40 to ~50% for insulin¹³ (Figure S13E and F) formed amyloid fibrils. These results indicate that native globular proteins exhibited a higher propensity for amyloid fibrillation when the contents of α-helix were generally higher (~20% to ~50%) than those of initially disordered proteins (~0 to ~40%) due to the alcohol-induced unfolding of native structures. Detailed information on helical states at the atomic and residue levels is needed in order to obtain a deeper understanding of the interplay between helical structures and aggregation propensities as well as aggregate types. Further case studies based on the phase diagram, which also includes detailed structural information, are required to identify a more general relationship among protein aggregation, structures, and solubilities. Finally, currently available phase diagrams will be applicable to the macroscopic and microscopic understanding of phase separation, which is a weak protein aggregation.⁵

CONCLUSIONS

Understanding of the molecular mechanisms of Aβ_1–40 aggregation under varying conditions is of significant importance for controlling amyloidogenesis of Aβ peptides and for developing effective therapeutic modalities. We herein demonstrated that Aβ_1–40 underwent diverse aggregation pathways depending on the initial structure and solubility at the residue and molecular level. We proposed a general concept for amyloid aggregation: partially helical structures are aggregation prone and amyloidogenic with a decrease in solubility, while largely helical structures maintain monomeric states due to high solubility. In addition to these microscopic molecular mechanisms, macroscopic phase diagrams of aggregation of Aβ_1–40 and several other amyloid proteins with information on initial structures further provided broad insights into supersaturation-limited amyloid generation of unfolded and structured proteins. Future studies on context-dependent aggregation based on microscopic and macroscopic viewpoints will improve our understanding of complicated Aβ_1–40 aggregation in vivo and proteopathy.

METHODS

Chemicals.

Lyophilized Aβ_1–40 purchased from Peptide Institute Inc. (Osaka, Japan) (purity ≥95.0%) was dissolved in 0.05% (w/w) ammonia solution at a concentration of 100 μM and stored at −80 °C, as described previously.⁵⁵ 3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) was obtained from Sigma-Aldrich. ThT was obtained from Wako Pure Chemical Industries, Ltd. TFE, HFIP, and other reagents were purchased from Nacalai Tesque (Kyoto, Japan).

Aβ_1–40 Oligomer Preparation.

Aβ_1–40 oligomers were prepared as described previously.^48,49 Briefly, we dissolved lyophilized Aβ_1–40 in 100% HFIP (v/v) at 2.5 mg mL⁻¹ (~580 μM) and incubated Aβ_1–40 at room temperature for 15 min. Samples were then diluted 10-fold with doubly deionized water. After a further incubation at room temperature for 15 min, samples were centrifuged at 14000g for 15 min to remove large aggregates.

Monitoring of Aβ_1–40 Aggregation in Microplates by the ThT Fluorescence Assay.

Sample solutions (200 μL) were transferred to each well of a 96-well microplate (Greiner-Bio-One, Tokyo, Japan), and a sealing film was affixed to prevent evaporation (PowerSeal Cristal View, Greiner-Bio-One). The microplate, set on a water bath-type ultrasonic transmitter with a temperature controller (Elestein SP070-PG-M, Elekon Sci. Inc., Chiba, Japan), was subjected to repetitive sonication pulses for 1 min at 9 min intervals from three directions. Temperature was maintained at 37 °C throughout the experiment. The frequency and power output of the ultrasonic waves were set to 17–20 kHz and 350 W, respectively. Unless otherwise stated, the following experimental conditions were used to produce amyloid fibrils: 10 μM Aβ_1–40, 10 mM sodium phosphate buffer (pH 7.5), 100 mM NaCl, 5 μM ThT, and alcohols at various concentrations. The context-dependent aggregation process of Aβ_1–40 oligomers in 2% HFIP (pH ~4) to other types of aggregates was first investigated at 37 °C with and without sonication and 10 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl. For more detailed information, another experimental condition with continuous shaking (425 cpm) was next used with and without 10 mM sodium phosphate buffer (pH 7.5) at the various concentrations of NaCl (0–2 M). ThT fluorescence intensity for all experiments was monitored using a microplate reader (MTP-810, Corona Electric Co. Ibaraki, Japan, or Synergy Neo2 Hybrid Multi-Mode, Biotek, VT, USA) with excitation and emission wavelengths of 450 and 490 nm, respectively.

Kinetic parameters of Aβ₁₋₄₀ fibrillation in TFE were obtained by fitting fibrillation curves to the sigmoid equation given below.

$$Y(t)=y_{\mathrm{i}}+m_{\mathrm{i}}t+\frac{y_{\mathrm{f}}+m_{\mathrm{f}}t}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left[-k\left(t-t_0\right)\right]}$$ (1)

where $y_{\mathrm{i}}+m_{\mathrm{i}}t$ and $y_{\mathrm{f}}+m_{\mathrm{f}}t$ are the initial and final baselines, respectively. $k$ and $t$ indicate the rate constant of elongation and time, respectively. $t_0$ is the half-time when ThT fluorescence reaches 50% of its maximum intensity. The lag time was calculated with the following relationship: $\mathrm{l}\mathrm{a}\mathrm{g}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}=t_0-2(1/k)$.⁹⁸

The apparent rate constant of the transition from protofibrils to mature amyloid fibrils in a water/HFIP mixture was obtained by fitting the transition time course to the following exponential function:

$$Y\left(t\right)=y_0+A{\mathrm{e}}^{-kt}$$ (2)

where $y_0$ is the offset and $A$ represents the amplitude of the change in the ThT intensity. $k$ is the rate constant for the protofibril–amyloid fibril transition and $t$ indicates the incubation time. All kinetic parameters were obtained using the averaged values of three independent data.

Monitoring of Aβ_1–40 Aggregation in Test Tubes by the ThT Fluorescence Assay.

Since sealing films are dissolved at high HFIP concentrations (>10%), Aβ_1–40 aggregation was monitored using test tubes in 10 mM sodium phosphate buffer (pH 7.5) containing 100 mM NaCl, 5 μM ThT, and 50% HFIP (v/v). The ultrasonic transmitter and experimental temperature were identical to those for measurements in microplates. Sample solution was transferred to a 0.5 cm light-path cuvette at the desired time points, and ThT fluorescence intensity was recorded using a Hitachi F4500 fluorescence spectrophotometer (Hitachi, Tokyo, Japan) with excitation and emission wavelengths of 450 and 490 nm, respectively.

CD Spectroscopy.

The far-UV CD spectra of Aβ_1–40 sample solutions were recorded using a Jasco J820 spectropolarimeter (Tokyo, Japan) equipped with a Peltier temperature cell holder controller. A cuvette with a light-path length of 1 mm was used. Sixteen scans of individual samples were accumulated and averaged. All spectra were expressed as the mean residue ellipticity, [θ] (deg cm² dmol⁻¹), after subtracting the signal from a solution without Aβ_1–40.

AFM and TEM.

AFM images were obtained using a Digital Instruments Nanoscope IIIa scanning microscope (Veeco Instruments Inc., NY, USA). Sample solutions with aliquots of 20 μL were spotted on freshly cleaved mica plates and incubated for 10–15 min. Samples were then gently rinsed three times with 20 μL of doubly deionized water to wash off salt crystals. Residual water was blown off with compressed air and then air-dried. TEM images were measured using a Bio-High voltage EM system (JEM-1400 Plus at 120 kV and JEM-1000 BEF at 1000 kV; JEOL Ltd., Tokyo, Japan), a monochromated Cs-corrected TEM (Mono ARM200, JEOL Ltd., Tokyo, Japan), and a Hitachi H-7650 transmission microscope (Hitachi, Tokyo, Japan) with a voltage of 80 kV, as reported previously.⁵⁵ For TEM analyses, we applied 4–5 μL of sample solution to a collodion-coated copper grid (Nisshin EM Co., Tokyo, Japan) and incubated it for 1 min. The remaining solution was then removed with filter paper, and 5 μL of doubly deionized water containing 2% (w/w) ammonium molybdate or 2% (w/w) uranyl acetate was spotted onto collodion-coated copper grids. After 1 min, the remaining solution was removed in the same manner.

Analytical Ultracentrifugation.

The size distribution of Aβ_1–40 oligomers in 2% HFIP (v/v) was investigated by sedimentation velocity measurements using a Beckman-Coulter Optima XL-A analytical ultracentrifuge (Beckman Coulter, FL, USA). Samples were centrifuged at 3000 rpm (700g) for 5 min to stabilize absorbance and temperature. The rotor speed was increased to 18 000 rpm (26000g). Absorbance data at 220 nm were recorded at 25 °C. Radial increments were set to 0.003 cm. The continuous scanning mode was selected with one scan every 15 min. We analyzed absorbance data using SEDFIT (http://www.analyticalultracentrifugation.com/default.htm), and parameters were calculated by SEDNTERP (http://bitcwiki.sr.unh.edu/index.php/Main_Page).

Dynamic Light Scattering.

Dynamic light scattering measurements were performed for an Aβ_1–40 oligomer solution containing 2% HFIP (v/v) using a DynaPro NanoStar instrument (Wyatt Technology Co., CA, USA) and a 100 μL plastic cuvette. Dust particles were removed from the sample before measurements by spinning down at 10000g for 5 min. The cuvette was placed in the sample chamber maintained at 25 °C. An average of scattering data was obtained using 20 separate scans collected over 20 s. We processed data using the manufacturer’s software (Dynamics; Wyatt Technology Co.).

MD Simulation.

The MD simulation systems were set up in reference to our experimental conditions. The Aβ_1–40 peptide (PDB ID: 1AML⁴⁵) was simulated in solution containing either 85% H₂O/15% TFE or 98% H₂O/2% HFIP. The topology files for the TFE and HFIP compounds were generated from the Automated Topology Builder.⁹⁹ SPC water model was employed. All simulations were performed using the GROMACS simulation suite version 5.0.7 and the Gromos53a6 force field. Aβ_1–40 was submerged in a cubic box with a minimum distance of 12 Å from the box boundary. The MD systems were subjected to a NaCl concentration of 100 mM and neutralized by adding requisite counterions. The equilibration of MD systems was achieved using NVT and NPT ensembles at 310 K and 1 bar using typical parameters described in our prior report.¹⁰⁰ A production MD run of 100 ns was considered for structural and dynamic interpretation. In order to generate an unfolded Aβ_1–40 conformation, a helical Aβ_1–40 molecule was simulated at 373 K for 100 ns and then gradually annealed to 310 K. The conformational transition of Aβ_1–40 from unfolded to helically folded states in the presence of either TFE or HFIP was thus investigated on a time scale of 100 ns. The protein structure visualizations were carried out using PyMOL (https://pymol.org/) and Discovery Studio Visualizer 3.5 (http://accelrys.com/products/collaborative-science/biovia-discovery-studio/). MD trajectory analysis was done using Visual Molecular Dynamics (www.ks.uiuc.edu/Research/vmd/).

Cytotoxicity Experiments.

Human neuroblastoma SH-SY5Y cells were purchased from the American Type Cell Collection (ATCC, VA, USA). Cells were maintained in media containing 50% minimum essential medium (Gibco, NY, USA) and 50% F12 (Gibco), supplemented with 5% (v/v) fetal bovine serum (Gibco) and 100 U/mL penicillin (Gibco). The cells were grown in a humidified atmosphere with 5% CO₂ at 37 °C. Cell viability was determined by the MTT assay. The cells were seeded in a 96-well microplate (15 000 cells in 100 μL per well). In order to identify the cytotoxicity of Aβ_1–40 amyloid fibrils, Aβ_1–40 (100 μM) in three different solvents, 0% alcohol, 15% TFE, or 10% HFIP, was incubated overnight at 37 °C. Formation of amyloid fibrils was confirmed using the ThT assay. Aβ_1–40 amyloid fibrils were diluted by 10-fold to obtain the final concentration of 10 μM and then treated with cells. After a 24 h incubation, MTT of 25 μM in PBS (pH 7.4) (Gibco) was incubated in each well, and the plate was incubated for 3 h at 37 °C. Formazan produced by the cells was solubilized by the addition of an acidic solution of N,N-dimethylformamide (50% (v/v)) (pH 4.5) and SDS (20% (w/v)) overnight at room temperature in the dark. The absorbance was measured at 600 nm using a SpectraMax M5e microplate reader (Molecular Devices, CA, USA). Cell viability was calculated relative to the absorbance of cells themselves. Error bars were calculated as standard errors from three independent experiments.

Enzyme-Linked Immunosorbent Assay.

A 96-well ELISA plate kit (Sumitomo Bakelite, Japan) was used. Each well was coated with 100 μL of an anti-amyloid oligomer antibody (1:100) (Merck, Germany) dissolved in the coating buffer supplied by the manufacturer. After wells were washed three times with washing buffer (10 mM phosphate buffer (pH 7.5) containing 100 mM NaCl and 0.05% Tween 20), 100 μL of 5.8 μM Aβ_1–40 monomers or oligomers formed with the same Aβ_1–40 concentration as well as 10 mM phosphate buffer (pH 7.5) containing 100 mM NaCl and 1% HFIP were added to the wells and incubated at 25 °C for 1 h. After wells were washed three times with washing buffer, bound Aβ_1–40 was detected with an anti-human Aβ antibody (Dako, CA, USA) and horseradish peroxidase-conjugated anti-mouse immunoglobulin antibody (1:1000) (Dako) followed by color development using 3,3′,5,5′-tetramethylbenzidine as the peroxidase substrate (Bio-Rad, CA, USA). Absorbance was measured at 450 nm using a SpectraMax M5 microplate reader (Molecular Devices, CA, USA). The absorbance of the well, which was not coated with the anti-amyloid oligomer antibody, was subtracted from each data point.

Intrahippocampal Injection of Aβ_1–40 Aggregates and Immunohistochemistry.

Aβ_1–40 oligomers and protofibrils formed in vitro were separately injected to wild-type mice. All experimental procedures were performed in accordance with approved animal protocols and guidelines established by the Korea Brain Research Institute (IACUC-2016–0013). Male C57BL6/N mice (8 weeks old and 25–30 g) were purchased from Orient-Bio Company (Gyeonggido, Korea) and housed in a pathogen-free facility with 12 h of light and dark per day at 22 °C. Mice were anaesthetized with avertin (2,2,2 tribromoethanol in 2-methyl-2-butanol and 250 mg/kg, i.p.) and placed in a stereotaxic apparatus (Stoelting, IL, USA). Oligomers, protofibrils, or vehicle controls were injected bilaterally into the hippocampus CA1 regions in a volume of 1.5 μL at a flow rate of 0.2 μL/min by using a Hamilton 10 μL syringe and a 26 G needle. The stereotaxical coordinates of the hippocampus CA1 injection site were −2.5 mm anterior/posterior, ±2.0 mm medial/lateral, and −2.0 mm dorsal/ventral from the bregma. At 24 h after injection, mice treated with injection of oligomers, protofibrils, or vehicles were perfused and fixed for immunohistochemistry.

Briefly, each brain section was rinsed with phosphate-buffered saline (PBS) three times and incubated with blocking solution containing 0.2% Triton X-100, 1% bovine serum albumin (BSA), and 4% goat serum in PBS for 90 min at room temperature. The brain sections were then washed twice with 0.5% BSA in PBST and incubated at 4 °C with anti-4G8 antibodies (1:200, Biolegend, CA, USA) overnight. The next day, the brain sections were washed with 0.5% BSA in PBST three times and incubated with Alexa 488-conjugated goat anti-mouse IgG (1:200, Life Technologies, CA, USA) for 90 min at room temperature. The brain sections were then rinsed with PBS three times, mounted on slide glass, and covered with DAPI-containing mounting solution (Vector Laboratories, CA, USA). Images of the stained tissues were captured using fluorescence microscopy (DMi8, Leica, Wetzlar, Germany), and fluorescence intensity was measured using ImageJ (NIH) software.