Single-Molecule Kinetic Observation of Antibody Interactions with Growing Amyloid β Fibrils

Abstract

Understanding the dynamic assembly process of amyloid β (Aβ) during fibril formation is essential for developing effective therapeutic strategies against Alzheimer’s disease. Here, we employed high-speed atomic force microscopy to observe the growth of Aβ fibrils at the single-molecule level, focusing specifically on their interaction with anti-Aβ antibodies. Our findings show that fibril growth consists of intermittent periods of elongation and pausing, which are dictated by the alternating addition of Aβ monomers to protofilaments. We highlight the distinctive interaction of antibody 4396C, which specifically binds to the fibril ends in the paused state, suggesting a unique mechanism to hinder fibril elongation. Through real-time visualization of fibril growth and antibody interactions combined with molecular simulation, this study provides a refined understanding of Aβ assembly during fibril formation and suggests novel strategies for Alzheimer’s therapy aimed at inhibiting the fibril elongation.

Introduction

Amyloid refers to abnormal protein structures that have a propensity to accumulate in various tissues within the body. Amyloidogenic proteins undergo dynamic conformational transitions, resulting in assembly and the formation of insoluble fibrils characterized by regularly stacked β sheets.¹ These fibrils can evoke cell dysfunction and tissue damage, contributing to the onset and progression of various diseases, including neurodegenerative disorders exemplified by Alzheimer’s disease and Parkinson’s disease.²

Amyloid β (Aβ) is a protein associated with Alzheimer’s disease, generated through the cleavage of an endogenous membrane protein known as amyloid precursor protein by two types of protein-degrading enzymes, β-secretase and γ-secretase, within the brain.³ The formation and morphological diversity of Aβ fibrils depend on various factors including Aβ species and environmental conditions such as pH, ion strength, pressure, and gravity.⁴ Despite such variabilities, Aβ fibrils exhibit consistent structural characteristics. They typically consist of cross-β structures, composed of two or more protofilaments, each featuring extended β-sheets arranged perpendicular to the direction of fibril growth.⁵

The fibrilization and deposition of Aβ as amyloid in the brain constitute fundamental events in Alzheimer’s disease; however, the precise mechanism underlying this process remains to be determined. Interestingly, a unique Aβ species, which was tightly associated with GM1 ganglioside, a glycosphingolipid abundant in neuronal membranes, was identified in the cerebral cortices of human brains exhibiting early pathological changes of Alzheimer’s disease.⁶ Furthermore, Aβ fibrilization was found to be accelerated on membranes containing GM1 ganglioside.⁷ Ever since then, ganglioside-bound Aβ has been extensively studied from various perspectives.⁸⁻¹⁰

Understanding the dynamic processes of Aβ assembly is crucial for designing amyloid inhibitors, including therapeutic antibodies. The Aβ assembly processes have predominantly been characterized through kinetic analysis utilizing thioflavin T (ThT) fluorescence as a probe to monitor amyloid fibril growth.¹¹ These analyses operate on the model that amyloid fibril formation encompasses nucleation, elongation, secondary nucleation, and fragmentation steps. While such kinetic analyses have provided insights into the mechanisms of action of anti-Aβ antibodies on fibril formation,¹² their macroscopic nature inherently limits the provision of structural details necessary for the design of amyloid inhibitors. Despite the invaluable microscopic structural information offered by NMR spectroscopy and molecular dynamics (MD) simulations regarding the early stages of Aβ assembly in both solution and membranous environments,^10,13 the visualization of fibril elongation and the mechanisms by which inhibitors suppress it remain challenging due to constraints imposed by spatial and temporal scales.

Real-time observation of single fibrils holds significant promise in bridging the gap between macroscopic and microscopic perspectives on fibril growth. A previous study using total internal reflection fluorescence microscopy has demonstrated that the growth of Aβ42 fibrils exhibits strong polarization, with distinct fast- and slow-growing ends.¹⁴ This polarized bidirectional growth pattern has also been corroborated by high-speed atomic force microscopy (HS-AFM), revealing intermittent growth characterized by a stop-and-go behavior in Aβ42 fibrils.¹⁵ In this context, MD simulations have indicated that Aβ42 at the growing end of an amyloid fibril adopts a β-hairpin conformation with less fluctuation compared to the flexible opposing terminus, suggesting the relevance of conformational dynamics at fibril ends in elongation processes.¹⁶

In our current study, we perform single-molecule kinetic observation of Aβ fibril elongation and its interactions with anti-Aβ antibodies by HS-AFM, thereby providing a theoretical model of fibril growth. Specifically, we highlight the unique property of a monoclonal antibody, 4396C, which was raised against the GM1-bound Aβ species isolated from human brain.¹⁷ Remarkably, 4396C is capable of strongly inhibiting fibril formation of Aβ on the GM1 membranes; moreover, it can suppress the elongation of fibrils even in the absence of GM1 ganglioside membranes. This suggests that it targets specific Aβ conformers at the growing end of the fibrils, which share a common epitope with the GM1-bound Aβ species. Therefore, elucidating the epitope details of 4396C will facilitate linking characteristic Aβ conformers during the early nucleation process on GM1 membranes with those crucial for fibril elongation, thereby providing valuable insights for the design and development of novel drugs aimed at counteracting Aβ fibrilization.

Results

HS-AFM Imaging of Elongation Occurring at the Ends of Aβ Fibrils

We initially monitored the growth of Aβ42 fibrils using preformed fibril seeds. HS-AFM visualized the elongation of Aβ42 fibrils, revealing a polarized, bidirectional growth with distinct fast and slow ends (Figure 1a,b, Supporting Figure S1, Movie S1). Analysis through kymographs depicts intermittent fibril growth at the fast end, characterized by alternating extension and resting phases (Figure 1c). These observations align qualitatively with prior reports, which were conducted under different experimental conditions,^15,18 with our particular focus directed toward the fast end of fibril growth. Histograms of the extension and resting phases of the Aβ42 fibril growth (Figure 1d) illustrate that the resting period lasted up to 80 s, with calculated time constants τ_ex = 4.24 ± 0.15 s and τ_r = 4.31 ± 0.37 s, respectively. Utilizing these time constants, we performed calculations to determine the fibril resting probability. This involved computing the ratio of resting time to the total fibril growth time, which revealed a resting probability of 50%.

Figure 1.

Intermittent growth of Aβ fibrils. (a) Time-lapse HS-AFM images showing the elongation of Aβ fibrils on a mica substrate at an imaging speed of 1 s/frame. Scale bar: 100 nm. Aβ42 was added at the final concentration of 0.5 μM to induce fibril growth. (b) Kymograph extracted along the fibril in (a). (c) The trajectory illustrates the growth of Aβ fibrils at the fast end. A stepwise fibril growth was observed, comprising extension (light red) and resting (light blue) phases. Upon zooming into the extension phase enclosed by the square, two distinct steps became evident: the elongation step (light red) and the pause step (blue hatched). Utilizing the time constant τ_r in (d), the difference in distance (absolute value) of ≤1 nm between frames persisting for the duration of the time constant or longer (over 4 s) was defined as the resting phase. (d) Statistical distributions of the extension (top left) and resting (top right) durations during Aβ fibril growth, acquired with an imaging speed of 1 s/frame. The histograms were fitted with a single exponential function to estimate the time constants for extension (τ_ex) and resting (τ_r) phases. The resting probability, defined as , was calculated, resulting in P_r = 0.5. The duration histograms for elongation (bottom left, n = 246) and pause (bottom right, n = 239) steps in the extension state of Aβ fibrils acquired with an imaging speed of 100 ms/frame. The cumulative frequencies (black plots) were fitted with a single exponential function to obtain τ_el and τ_p. The pause probability, defined as , was calculated, giving P_p = 0.74. The final concentration of Aβ42 used was 0.2 μM.

Operating at an imaging speed of 100 ms/frame, HS-AFM facilitated detailed examination of fibril growth, offering enhanced spatiotemporal resolution. The HS-AFM imaging allowed us to closely scrutinize the extension phase of the kymograph, depicting repeated periods of elongation and pause steps. We quantified the duration of the elongation and pause of the Aβ fibrils, with the time constants τ_el and τ_p representing the average duration of elongation and pause, respectively (Figure 1d). Utilizing these time constants, we defined the pause probability of the fibrils as the ratio of pause time to total extension time, revealing a pause probability of 74%. Combining the resting and pause phases revealed that the probability of “stop” during the fibril growth process was 87%, over six times more frequent than elongation.

Furthermore, this approach provided detailed insights into the growing end of a fibril, composed of two protofilaments extending in a stepwise fashion (Figure 2a, Supporting Movie S2). High-resolution cryo-electron microscopy (cryo-EM) has revealed that Aβ42 fibrils are composed of two LS-shaped protofilaments. Each protofilament features an L-shaped segment spanning from the N-terminus to the central region (Asp1-Asp23) and an S-shaped segment extending from the central region to the C-terminus (Leu17-Ala42) (PDB: 5OQV).¹⁹ Furthermore, hydrophobic interactions within the C-terminal region (Val39-Ala42) stabilize the interface between the two protofilaments. To estimate the step size in fibril elongation, we generated pseudo-AFM images simulating the addition of monomeric Aβ42 to the end of the cryo-EM structure of Aβ42 fibril, indicating that one monomer binding corresponded to an extension distance of 1.0 nm (Figure 2b,c). The distribution of step size exhibited a primary mode of 1.0 nm under our experimental conditions (Figure 2d), indicating that the fibril elongation progressed by the addition of one Aβ42 molecule per step of each protofilament growth. Statistical analysis revealed that alternating elongation of the two protofilament ends occurred in over 80% of cases (Figure 2e). In other minor cases, elongation occurred successively within the same protofilament. During the paused state, more than 80% of cases showed evenly matched edges of the two protofilaments (Figure 2f).

Figure 2.

Microscopic processes of Aβ fibril elongation at the fibril end. (a) HS-AFM images capture the elongation process of Aβ fibril, illustrating a sequential elongation where only one end of the two protofilaments constituting the Aβ fibril elongated initially (termed “odd edge”), succeeded by elongation at the other end, leading to the alignment of ends (referred to as “even edge”). The imaging was performed at an imaging speed of 100 ms/frame. Scale bar: 10 nm. For comparative analysis of tip positions, a dashed line connecting the ends of the two protofilaments in the initial frame is shown in yellow in each frame. Additionally, red arrows indicate the protofilaments that have elongated compared to the previous frame. (b) A stereoscopic model of the Aβ fibril (PDB: 5OQV) is presented, with monomeric Aβ alternately added to the ends, simulating the fibril elongation by the smallest unit. (c) Pseudo-AFM image based on the structure in (b), with monomer Aβ binding corresponding to a size of 1.0 nm. The dashed line connecting the ends of the two protofibrils in the leftmost frame is replicated in the other frames for comparison. Scale bar: 10 nm. (d) Histogram displaying the interframe elongation distances of Aβ fibrils. The elongation distance for each frame, calculated as the difference in the position of the fibril ends, is counted (n = 182) at an imaging speed of 100 ms/frame. The final concentration of Aβ42 was 0.2 μM. The histogram could be fitted with the sum of two Gaussian distributions (black line), with a major peak at 1.0 nm and a subpeak at 1.5 nm. (e) The proportion of alternating elongation of the two protofilaments. (f) Proportion of even and odd edges in the pause state.

HS-AFM Imaging of Interactions of Aβ Fibrils with Monoclonal Antibodies

We conducted a comparative analysis of Aβ42 fibril growth in the absence and presence of three anti-Aβ antibodies, 6E10, 4G8, and 4396C.^17,20 HS-AFM data revealed distinct binding patterns of these antibodies to the Aβ fibrils. Both the 6E10 and 4G8 antibodies were primarily bound to the lateral sides of the Aβ fibrils. However, 6E10 exhibited prolonged binding and eventually covered the entire surfaces of the fibrils. The average number of 6E10 antibodies bound to the lateral side of the fibril was approximately seven times higher than for 4G8 (Figure 3a,b, Supporting Movies S3 and S4). This difference in binding is likely due to the fact that the 6E10 epitope (Asp1-Lys16) within the N-terminal segment is more exposed on the fibril surface compared to the 4G8 epitope (Leu17-Val24). In contrast, 4396C specifically bound to the ends of the fibrils but not to their lateral side (Figure 3c,d, Supporting Movies S5 and S6).

Figure 3.

Visualization of the interaction between Aβ fibrils and antibodies. HS-AFM images of elongating Aβ fibrils after adding either (a) 6E10 or (b) 4G8. HS-AFM images captured at 10 and 15 min after adding antibodies revealed distinct binding patterns. In the case of 6E10, the antibody bound to the lateral side of the Aβ fibrils. For the experiments involving 6E10, the final concentrations of Aβ42 and 6E10 were 0.3 μM and 16 nM, respectively, while in the 4G8 experiments, the final concentrations of Aβ42 and 4G8 were 0.5 μM and 32 nM, respectively. Scale bar: 30 nm. (c) HS-AFM images taken at 0 min (top left) and 1.6 min (top right) after the addition of 4396C. Scale bar: 30 nm. (d) HS-AFM snapshots at 2.5, 5, 7.5, and 10 min of events where 4396C continuously bound to the end of Aβ fibrils for several minutes without dissociation within the observation time, thereby inhibiting elongation. Scale bar: 20 nm. The final concentrations of Aβ42 and 4396C were 0.5 μM and 32 nM, respectively. The white arrows indicate the locations of antibodies.

To investigate whether the binding of 4396C depends on the number of fibril edges, we examined the effects of fibril fragmentation on the reactivities of these monoclonal antibodies using a ThT assay. The application of ultrasound pulses led to the fragmentation of Aβ fibrils into smaller pieces as confirmed by negative-stain transmission electron microscopy (TEM) (Figure 4a). This fragmentation was accompanied by an increase in the number of fibrils with growing ends, serving as active seeds for fibril growth, as indicated by ThT fluorescence enhancement (Figure 4b). Unlike 6E10 and 4G8, the 4396C antibody exhibited greater reactivity with shorter fibril fragments than the original longer fibrils (Figure 4c). These findings suggest that the binding of 4396C depends on the number of fibril edges, consistent with the earlier HS-AFM data, indicating that the epitope recognized by 4396C is located on the Aβ fibril edge. Notably, fibril elongation did not occur while 4396C was bound to its growing end, a phenomenon not observed with 6E10 and 4G8, which bind to the lateral surfaces of the fibrils (Figure 3). Consequently, we will focus on further quantitative analysis of the interactions between 4396C and Aβ fibrils.

Figure 4.

Reactivity of 4396C antibody to the active ends of fibrils. (a) TEM images of Aβ fibrils after various durations of ultrasound pulse application, illustrating the fragmentation process. (b) ThT fluorescence assay of Aβ42 fibrilization in the absence (gray) and presence of 0.2% (v/v) ultrasound-induced fragments. The color codes for the number of pulses are given on each row. (c) Dot blot assay of the ultrasound-induced fragments with monoclonal anti-Aβ antibodies: 4396C (left), 6E10 (middle), and 4G8 (right).

The ThT assay provides a macroscopic perspective on the inhibitory effect of 4396C on Aβ fibril formation. The data revealed that 4396C had minimal impact on the lag time for the increase in fluorescence intensity but significantly reduced the steepness of the curve (Figure 5), indicating that the antibody inhibited fibril elongation rather than the primary nucleation process. As mentioned earlier, the progression of Aβ42 fibril growth is characterized by occasional pauses. We explored whether 4396C binds to the Aβ42 fibril edge more frequently during the elongation or pause phases. In many instances, when 4396C bound to the fibril end, there was an observable increase in fibril image length corresponding to the size of the antibody, and fibril growth ceased both before and after binding (as illustrated in Figure S2). Consequently, the HS-AFM data revealed that 87% of the binding events of 4396C occurred at least 0.1 s after the spontaneous cessation of fibril elongation (Figure 6a), indicating that 4396C preferentially binds the fibril end during the pause state. Statistical analysis of HS-AFM data suggested that the pause period of fibril extension was significantly extended in the presence of 4396C, averaging 54.6 s compared to an average of 5.5 s in the absence of 4396C. Moreover, 4396C often remained bound to the fibril edge for more than 100 s, effectively arresting the fibril elongation (Figure 6b). Increasing the 4396C concentration led to a more potent inhibition of overall Aβ fibril growth (Figure S3).

Figure 5.

Inhibitory effect of 4396C on fibril formation. (a) ThT fluorescence as a function of time for reactions starting with 1 μM Aβ42 in the absence and presence of 4396C. (b) The data are shown in normalized form. The color codes representing the antibody concentrations in μM are given on each row.

Figure 6.

Impact of 4396C binding on pause duration in Aβ fibril elongation. (a) Time and frequency from the cessation of fibril growth to antibody attachment (n = 159). Cases where an antibody bound within 0.1 s (1 frame) after fibril growth stopped were excluded from the analysis (n = 21, 13% of the total). The duration until antibody binding comprises two components: instances where antibodies attach to the fibril incidentally (short-time component, τ₁ = 1.01 ± 0.07 s) and cases where antibodies specifically bind after recognizing the fibril in a paused state (long-time component, τ₂ = 5.26 ± 0.63 s). (b) Logarithmic graph of pause durations and count numbers during Aβ fibril elongation in the presence of 0.5 μM Aβ42. Measured pause durations are shown both in the absence (gray, average 5.5 s, n = 150) and presence of 4396C at a final concentration of 16 nM (orange, average 55.6 s, n = 85).

Discussion

The formation of amyloid fibrils is a complex process that involves nucleation, elongation, and secondary nucleation. Numerous experimental and theoretical studies have explored the mechanisms underlying fibril elongation. An earlier kinetic study utilizing radiolabeled Aβ40 suggested a two-step process for fibril elongation: initially, a monomer binds reversibly to the growing end of the fibril on a fast time scale, followed by a conformational transition on a slower time scale.²¹ This transition confers a higher affinity for the fibril. This is often referred to as the “dock-lock mechanism”. Subsequent advancements in microscopic techniques have enabled real-time observation of fibril growth at the single-fibril level, extending beyond Aβ to encompass various amyloidogenic proteins such as glucagon, α-synuclein, and amylin.²² Cumulative data reveal a fundamental characteristic of amyloid fibril growth: elongation occurs in a start-and-stop fashion, suggesting a dynamic transition between active (growing) and inactive (blocked) states, with the fibril kinetically trapped in the latter state due to inherent conformational attributes at its edge.

In this study, we employed HS-AFM to observe the elongation of Aβ42 fibrils, focusing on the growing end of the fibril composed of two protofilaments. Our real-time observation unveiled that, in most cases, the two protofilaments aligned their elongations by taking one step forward alternately (Figure 2). When the ends of the two protofilaments were not aligned (in an “odd edge” state), priority was given to aligning them into an “even edge” state. In the cross-β structure of Aβ42 fibrils, each protomer is stacked via hydrogen bonding (longitudinal interaction), forming a β-sheet in one protofilament, while simultaneously making hydrophobic contacts with the counterpart protomer in the other protofilament (lateral interaction). This dual interaction accommodates a protomer at the fibril end, explaining the preference for an even edge, rather than one protofilament elongating prominently.

The alternating elongation of the two protofilaments indicates that the two protomers are not equivalent even at the even edge, as the protomer incorporated into the fibril ahead tends to be the first to accept the new monomer in the next round (Figure 7). This can be analogized to the dock-lock mechanism, where a protomer bound to the protofilament end transforms into an active conformation, ready to accept the next protomer along the same protofilament. This conformational transition is a slower process compared to the time scale of monomer binding to the odd edge, which can take place prior to the completion of the conformational transition of the protruding protomer, especially if only the nonprotruding protomer is already in an active conformational state and ready for longitudinal interaction. We did not observe the reversible step, i.e., dissociation of monomer from the fibril end, which was originally underlined in the dock-lock mechanism, presumably due to a short lifetime of the “dock-but-unlock” state compared to the time scale of HS-AFM observation.

Figure 7.

Schematic model of Aβ fibril elongation. The diagram illustrates the alternating elongation mechanism of two protofilaments, emphasizing the nonequivalence of protomers at the edge of the fibril (depicted at the bottom). The leading protomer, integrated into the fibril, exhibits a predisposition to readily accept a new monomer in subsequent rounds. The bound protomer (depicted as a “white pentagon with a curved side”) at the end of the protofilament undergoes a conformational transition via an intermediate state (illustrated as a “pink pentagon with a curved side”) to an active state (illustrated as “red arrowhead”) that can accommodate the next protomer (represented as “white pentagon with a curved side”). This process is a relatively slower event compared to monomer binding at the odd edge. At the even edge of the fibril, protomers at the ends of the protofilaments slowly undergo conformational transitions with a time delay, resulting in two unorganized protomers aligned side by side for a certain period, occasionally causing a misalignment (indicated by the “pink hexagon”), which halts the fibril elongation until this misalignment is rectified (as indicated at the top). During the resting and pause phases, the fibril ends are trapped in a distinctive conformational state, characterized by an epitope recognized by monoclonal antibody 4396C.

We conducted a Monte Carlo (MC) simulation of fibril elongation that incorporated the slow conformational transition from the inactive to an active state of the protomer after binding to the protofilament end (Figure 8). Our simulation, based on this model (Model 1), reproduced an alternating elongation of the two protofilaments depending on the time constant of the conformational transition (Figure 8a–c, Supporting Movie S7). Although the specific mechanism of the conformational change remains elusive, it is plausible to hypothesize that a protomer newly bound to the fibril edge is primed for (or has already achieved) the lateral hydrophobic interaction but is not yet ready for the longitudinal interaction, i.e., the extension of the β-sheets. In mature Aβ42 fibrils, lateral interactions are primarily mediated through the C-terminal hydrophobic segments. Therefore, we speculate that the slow conformational transition involves a process of conformational organization of the remaining segment to facilitate synergistic hydrogen bonding, thus extending the β-sheets.

Figure 8.

Comparative analysis of Aβ fibril growth in lattice Monte Carlo simulation. Lattice Monte Carlo simulations of Aβ fibril elongation conducted using Model 1 (panels a–c) and Model 2 (panels d–f). (a, d) The progression of two protofilaments length over time. The orange and purple lines correspond to the respective protofilaments of an amyloid fibril. (b, e) The difference in length between the two protofilaments. (c, f) Snapshots taken from the Monte Carlo simulations at specific time steps, indicated by the corresponding dots in (a) and (d). The images represent amyloid fibril (red) and monomer (white) only in the yz plane.

Another notable aspect of amyloid formation is the intermittent pause in fibril growth, which cannot be explained simply by the aforementioned model. Antibody 4396C targets the edge of Aβ42 fibrils in the pause state, wherein most fibrils exhibit an even edge. This observation suggests that the fibril edge in the pause phase is trapped in a distinct conformational state, which bears an epitope recognized by 4396C (Figure 7). This monoclonal antibody was originally raised against Aβ42 species bound to ganglioside GM1, which was isolated from the cerebral cortices with early pathological changes of Alzheimer’s disease.¹⁷ Our recent solid-state NMR study has unveiled that Aβ molecules assemble into a nonfibrillar structure on GM1 ganglioside clusters, which exhibit reactivity with 4396C and feature a double-layered antiparallel β-structure.¹⁰ Within this structure, a solvent-exposed hydrophobic layer of the first β strands (referred to as β1 layer), composed of the Lys16-Leu17-Val18-Phe19-Phe20 segments in an antiparallel arrangement, serves as a catalytic surface triggering Aβ fibril formation. Moreover, the β1 layer is suggested to present a conformational epitope recognized by 4396C. The cross-reactivity of 4396C suggests the presence of a shared conformational epitope between the Aβ fibril edge in the paused state and the β1 layer of Aβ assemblage on GM1 ganglioside clusters.

At the fibril edge, each protomer at the protofilament end undergoes a slow conformational transition, initiated with a time lag. Consequently, at the even edge, two protomers not yet organized in conformation can align alongside each other for a certain duration. Under such circumstances, the unstructured segments of the two protomers could interact with each other, occasionally leading to misalignment. This could hinder fibril elongation, resulting in the emergence of the conformational epitope recognized by 4396C. We conducted another MC simulation of fibril elongation based on the hypothetical model described above (Model 2) (Figure 8d–f, Supporting Movie S8). The result successfully reproduced the alternating elongation of two protofilaments punctuated by intermittent pauses, while maintaining an even edge state for a period, supporting the model. Assuming that each protomer at the end of a protofilament undergoes a conformational transition without a time lag, no alternating elongation or intermittent pausing of the two protofilaments was reproduced (Model 3) (Supporting Figure S4, Movie S9). Thus, Model 2 best reflected the experimental results, strongly suggesting that during the pause phase, the fibril edge is trapped in a distinct conformational state, characterized by a specific time constant for the transition (Supporting Figure S5). As demonstrated by the HS-AFM data, Aβ fibrils are often kinetically trapped. Antibody 4396C binds to the edge of the fibril in this state, thereby delaying or virtually halting the fibril elongation (Figures 3 and 6). Although the structural entity of the 4396C epitope remains speculative, considering the structural data of GM1-bound Aβ,¹⁰ it is plausible that the β1 segments of the two Aβ protofilaments in an antiparallel arrangement form the epitope.

Our findings in this study offer new insights into the general mechanisms underlying amyloid fibril elongation and present a novel strategy for terminating Aβ fibril elongation by targeting the fibril ends, thus potentially slowing down the progression of Alzheimer’s disease. While this study focuses on the fibril elongation process, elucidating the primary and secondary nucleation processes, along with environmental effects and familial mutational effects,²³ is also important for a comprehensive understanding of the amyloid formation mechanisms. Factors such as liquid–liquid phase separation and membrane microdomains should also be considered in these complex processes,^8,24 and single-molecule observation techniques like HS-AFM can be expected to play a vital role in advancing this line of research, which will open up new possibilities for the prevention and treatment of Alzheimer’s disease.

Experimental Section

Preparation of Aβ42

Synthetic Aβ42 peptide, purchased from Toray, was employed in subsequent experiments after purification, during which the high-molecular-weight oligomer fraction was eliminated through chromatographic separation using a Superdex 75 Increase 10/300 column (Cytiva), following a protocol outlined in the literature.²⁵

Preformed fibrils were prepared by incubating 100 μM Aβ42 in a 10 mM sodium phosphate buffer at pH 7.4 at 37 °C. ThT fluorescence was monitored using a portion of the preformed fibrils to confirm fibril formation.

Antibodies

The experiments utilized mouse monoclonal antibodies 6E10 (BioLegend, #803002, Lot. B217405) and 4G8 (BioLegend, #800702, Lot. B220782), targeting amino acid residues Asp1–Lys16 and Leu17–Val24, respectively, of human Aβ. Mouse monoclonal antibody 4396C was obtained through contract synthesis (MBL, Lot. 1B505).

HS-AFM

HS-AFM imaging of Aβ fibrils was conducted using a laboratory-built apparatus²⁶ operated in tapping mode. We used a miniaturized cantilever with dimensions of 6–7 μm long, 2 μm wide, and 90 nm thick manufactured by Olympus (AC7). The mechanical properties of this cantilever had a resonant frequency of about 600 kHz in solution, a Q value of about 2, and a spring constant of about 0.2 N/m. The cantilever’s free oscillation amplitude was set to 3–4 nm and maintained at 60–90% of the free oscillation amplitude for feedback control. All observations were conducted at room temperature, approximately 25 °C. Aβ fibrils were diluted to 6.7 μM (monomer equivalent) in observation solution (10 mM potassium phosphate buffer, 80 mM KCl, pH 7.4), fragmented using a tip sonicator, and then adsorbed onto a mica substrate. The mica substrate, consisting of a disc-shaped mica of φ1.5 mm (from Furuchi Chemical Co., Ltd.) fixed onto a cylindrical glass stage (φ1.5 mm, height 2 mm from Japan Cell Co., Ltd.), was prepared with acryl adhesive. Adsorption of Aβ fibrils onto the mica substrate involved dropping 2 μL of 6.7 μM sample solution and incubating for 5 min. Subsequently, the substrate was washed with 60 μL of observation solution to remove unbound Aβ fibrils. HS-AFM observations were conducted in 80 μL of observation solution, and the concentrations of Aβ42 and antibodies added to the observation solution during observation were adjusted as necessary for each experiment. The concentration of Aβ42 injected during the observation of antibodies was maintained at a constant 0.5 μM. Histogram fitting analysis was conducted using the regression analysis standard in Igor Pro (WaveMetrics Inc.), with the standard error of the fitting serving as the attached error to each time constant.

TEM

Preformed Aβ42 fibrils were subjected to sonication using a probe-type sonicator (UD-201, TOMY), followed by negative stain EM analyses. Briefly, the specimens were stained with 2% uranyl acetate on the grid and visualized using a transmission electron microscope (JEM1010, JEOL) operated at an accelerated voltage of 80 kV.

Dot Blot Assay

Blotting was conducted using either 4396C or two different types of Aβ sequence-specific antibodies (6E10, 4G8, Millipore). Aβ42 fibril aliquots were spotted onto a nitrocellulose membrane (0.2 μm; BioRad), followed by drying and blocking with Blocking One (Nacalai Tesque) before immunodetection. 4396C was diluted by 1/200, while 6E10 and 4G8 were diluted to 1/500 as per the manufacturer’s instructions. Subsequently, an HRP-conjugated secondary antibody (Anti-Mouse IgG (H+L), HRP conjugate, #W4021, Promega) was added, and chemiluminescence detection was carried out using LAS3000 mini (FUJIFILM).

ThT Assay

Aβ42 peptides were initially dissolved in 6 M GuHCl and purified using a Superdex 75 Increase 10/300 column (Cytiva) at a flow rate of 0.5 mL/min in 20 mM sodium phosphate buffer (pH 8) with 200 μM EDTA and 0.02% NaN₃ to eliminate potential aggregated species. After purification, the obtained monomer was diluted to the desired concentration in buffer and supplemented with 20 μM ThT from a 2 mM stock solution. For experiments involving the addition of 4396C, the concentration of Aβ42 was maintained at 1 μM, with 4396C added at concentrations ranging from 5 to 200 nM. In experiments concerning ultrasound-induced fragments, fibril fragments were generated by sonication of 10 μM Aβ42 fibrils. The final concentration of fibrils, in monomer equivalents, was considered equal to the initial concentration of the monomer. These ultrasound-induced fibril fragments were then added to 5 μM of freshly prepared monomeric Aβ42 to achieve a final fibril concentration of 0.2%.

All samples were meticulously prepared in low-binding Eppendorf tubes on ice, with careful pipetting techniques to prevent the introduction of air bubbles. Subsequently, each sample was dispensed into multiple wells of a 96-well half area, low-binding polyethylene glycol coating plate (Corning 3881) with a clear bottom, at 100 μL per well.

The kinetic assays were initiated by incubating the 96-well plate at 37 °C under quiescent conditions in a plate reader (Infinite 200Pro, TECAN). The ThT fluorescence was measured through the bottom of the plate using a 430 nm excitation filter and a 485 nm emission filter. ThT fluorescence was monitored for three repeats of each sample.

Monte Carlo Simulation of Lattice Amyloid Model

Each Aβ42 molecule was represented as a single particle within a three-dimensional lattice with a lattice spacing of σ. These Aβ42 molecules could exist in one of three states: monomeric, fibrous, or intermediate. In Figure 9, a monomeric Aβ42 molecule was depicted as a white sphere, a fibrous molecule as a red sphere, and the intermediate state as a pink sphere.

Figure 9.

Lattice amyloid model for Monte Carlo simulation of fibril growth. (a) Monomer molecules (white sphere) and fibrous molecules (red sphere) in the yz plane in a three-dimensional lattice with a lattice spacing of σ. The amyloid fibril, depicted as red spheres arranged in two vertical columns, extends along the z-axis. (b) Interaction direction of a fibrous molecule in the left and right columns indicated by red arrows. (c) A monomer molecule (white sphere) bound to the end of amyloid fibril undergoes an intermediate state (pink sphere) and is transformed into a fibrous molecule (red sphere). This transformation takes time τ_lag.

The potential energy between two monomer molecules separated by r is given by

That between two fibrous molecules is given by

where ϵ represents the depth of potential energy when two fibrous molecules are positioned adjacent to each other. Note that the intermolecular distance r takes discrete values due to the arrangement of molecules on the lattice. The amyloid fibril was depicted as spheres organized in two columns (corresponding to two protofilaments), extending upward along the z-axis. The interaction direction between a fibrous molecule and a monomer molecule was defined, as shown in Figure 9b. A fibrous molecule on the left side exhibited an upward and rightward interaction direction, while one on the right side showed an upward and leftward direction. The potential energy between a fibrous molecule and a monomeric molecule in these directions is given by

The potential energy in the other direction is the same as that between monomers:

The intermediate state serves as a transitional phase between the monomeric and fibrous states. When a monomeric molecule attaches to the top of the protofilament, it transitions into the intermediate state. The conversion of intermediate molecules to fibrous molecules requires τ_lag, as illustrated in Figure 9c. Upon departing from the protofilament, intermediate and fibrous molecules return to monomers. The potential energy between a fibrous molecule and an intermediate molecule is given by

where t represents the time elapsed since the molecule transitioned from the monomeric state to the intermediate state. The potential energy between an intermediate molecule and a monomeric molecule in the interaction direction is given by

The potential energy between two molecules in the intermediate state is contingent upon the time since their transition to this state. Let t₁ denote the time elapsed since molecule 1 entered the intermediate state, and t₂ denote the time elapsed since molecule 2 entered the intermediate state. The potential energy between them is given by

These equations imply that the interaction between monomeric molecules gradually intensifies, leading to the transition into fibrous molecules. In the actual elongation process of Aβ amyloid fibrils, growth often experiences pauses. To replicate this phenomenon, when two molecules in the intermediate state aligned at the even edge, the interaction between an intermediate and monomeric molecule was set as the same as the interaction between monomeric molecules (i.e., u_II(r) = 0) for a duration of τ_stop with a probability of P_stop.

We conducted Monte Carlo simulations of the lattice amyloid fibril using the following parameters. The cubic simulation box was prepared with a side length of 20 σ. Within the box, ten fibrous molecules were placed in two columns. Additionally, 190 monomeric molecules were randomly distributed around the amyloid fibril, resulting in a total of 200 molecules. The temperature was set at k_BT = 0.1ϵ, where k_B represents the Boltzmann constant. Three Monte Carlo simulations were performed under different conditions as described below.

In Model 1, the transition of a monomeric molecule into a fibrous molecule required τ_lag = 100 steps. Model 2 introduced a stochastic and temporal pause of fibril growth. Model 3, on the other hand, denotes the absence of lag time: immediately after a monomeric molecule binds to the fibril from an upward direction, it transforms into a fibrous molecule at the subsequent time step. The time evolution of the Monte Carlo simulations was governed by the Metropolis algorithm.²⁷ The simulations were continued until sufficiently long amyloid fibrils were formed, with 5,000 steps in Simulation 1, 12,000 steps in Simulation 2, and 2,500 steps in Simulation 3.

Glossary

Abbreviations

Aβ

amyloid β

cryo-EM

cryo-electron microscopy

HS-AFM

high-speed atomic force microscopy

MC

Monte Carlo

MD

molecular dynamics

TEM

transmission electron microscopy

ThT

thioflavin T

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c08841.

• Polarized, bidirectional growth of Aβ fibrils, timing of 4396C binding to the end of Aβ fibrils, inhibition of Aβ fibril growth dependent on 4396C concentration, lattice Monte Carlo simulations in Model 3, and schematic model of Aβ fibril elongation with lattice Monte Carlo simulation parameters (PDF)

• HS-AFM observation of Aβ42 fibril elongation (MP4)

• HS-AFM observation of growing fibril end (MP4)

• HS-AFM observation of Aβ42 fibril with 6E10 (MP4)

• HS-AFM observation of Aβ42 fibril with 4G8 (MP4)

• HS-AFM observation of Aβ42 fibril with 4396C (MP4)

• HS-AFM observation of 4396C bound to fibril end (MP4)

• Lattice MC simulation of Aβ fibril elongation based on Model 1 (MP4)

• Lattice MC simulation of Aβ fibril elongation based on Model 2 (MP4)

• Lattice MC simulation of Aβ fibril elongation based on Model 3 (MP4)

Author Contributions

^# M.Y.-U., Y.K., and S.M. contributed equally.

This work was supported in part by JSPS KAKENHI (Grant Number JP19K07041 to M.Y-U., JP21K06118 to H.O., JP21H00393, JP21H01772 to T.U.), by JST PRESTO (Grant Number JPMJPR22AC to M.Y.-U.), by MEXT Promotion of Development of a Joint Usage/Research System Project: Coalition of Universities for Research Excellence Program (CURE) (Grant Number JPMXP1323015482), by Research Support Project for Life Science and Drug Discovery (Basis for Supporting Innovative Drug Discovery and Life Science Research (BINDS)) from AMED (Grant Number JP24ama121005), by Grant-in-Aid for Research in Nagoya City University (Grant Numbers 2212008, 2222004, and 2412012 to M.Y.-U.), the Naito Foundation Grant for Studying Overseas to M.Y.-U., by Joint Research of the Exploratory Research Center on Life and Living Systems (ExCELLS program No. 22EXC338, 23EXC305, and 24EXC329 to K.Y., 18-101 to T.U.), and by ExCELLS Project Research (ExCELLS program No. 22EXC601 to K.K. and T.U.).

The authors declare no competing financial interest.