**4. HS-AFM Observation of A**β**42 Aggregates on the O**ff**-Pathway**

### *4.1. HS-AFM Observation of Dissociating A*β*42 Aggregates*

HS-AFM visualizes the structural dynamics of non-fibrous Aβ42 aggregates. The HMW fraction, which was obtained by gel filtration, mainly contained globular aggregates [93]. They were classified into three different types of structural dynamics after the addition of 0.1 M sodium chloride: (1) gradually dissociating aggregates, (2) aggregates with unchanging sizes, and (3) stepwise dissociating aggregates (Figure 6a–f) [93]. The type I aggregates exponentially decreased in size (measured as the height from the HS-AFM stage) and then reached certain non-zero sizes, which suggests that the type I aggregates were composed of at least two parts with different dissociation constants (Figure 6d) [93]. The time course of the type I size, h(t), can be expressed as Equation (1):

$$\mathbf{h}(\mathbf{t}) = \mathbf{h}\_{\text{n0}} \cdot \exp(-\mathbf{k}\_{\text{n0}} \cdot \mathbf{t}) + \mathbf{h}\_{\text{k0}} \cdot \exp(-\mathbf{k}\_{\text{k0}} \cdot \mathbf{t}) \tag{1}$$

where ha0 and ka0 are the initial height and dissociation constant of part a, and hb0 and kb0 are the initial height and dissociation constant of part b. When the dissociation rate is extremely low (kb0 - 0), Equation (1) can be rewritten as Equation (2):

$$\mathbf{h}(\mathbf{t}) = \mathbf{h}\_{\text{a0}} \cdot \exp(-\mathbf{k}\_{\text{a0}} \cdot \mathbf{t}) + \mathbf{h}\_{\text{b0}}.\tag{2}$$

Equation (2) fitted the time courses of the type I aggregates well (Figure 6d) [93]. These analyses indicated that the type I aggregates formed during the sample preparation in 10 mM sodium phosphate, pH 7.4, and were then dissociated by the equilibrium shift after the addition of 0.1 M sodium chloride. For the relationship between the type I, II, and III aggregates, there were at least two possibilities: (1) they were distinct assemblies or (2) type II and III correspond to the slow dissociating part (the part b) of the type I aggregates [93]. The fast dissociating parts were completely dissociated before the aggregates appeared in the observation area, which may correspond to the type II aggregates (Figure 6b,d) [93]. We should also note that the tapping AFM probe may accidentally break the type II aggregates, resulting in a stepwise decrease in the aggregate size and thus generate type III aggregates (Figure 6f) [93].

**Figure 6.** *Cont*.

**Figure 6.** HS-AFM observation of non-fibrous aggregates in high molecular weight (HMW) Aβ42 incubation. (**a**) HS-AFM images of HMW Aβ42 incubation at the indicated time after the addition of 0.1 M NaCl. A closed triangle in the highlighted dashed box indicates a representative short fibril in HMW incubation. Bar, 300 nm. (**b**,**c**) HS-AFM image of two types of spherical aggregates in HMW incubation: type I gradually decreased in height (**b**); type II maintained its height (**c**). Bars, 50 nm. (**d**–**f**) Five representative time courses of height of type I (**d**), type II (**e**), and type III (**f**) aggregates in HMW incubation after an addition of 0.1 M NaCl. Each of the trajectories for type I is shown with the fitting curve of Equation (2). The dashed line corresponds to Equation (2), with the median values obtained from the best-fit values for individual type I aggregates. The type III aggregates show the sudden decrease in their heights at the time indicated by closed triangles. Different colors correspond to different single aggregates. Reproduced from [93].

#### *4.2. Characterization of Aggregation Pathway by Statistical Analysis*

HS-AFM analysis of the frequency of appearance of specific amyloid aggregates identified the position of individual aggregates during the molecular process. Figure 7a shows the time course of the cumulative number of fibrils appearing in the observation area after the addition of 0.1 M sodium chloride to the LMW or the HMW Aβ42 fractions [93]. The graph shows an upward trend in which fibrils appeared and accumulated. Although the LMW appeared throughout the time period, the HMW began to increase in the number of fibrils 20 to 30 min after the incubation. This tendency can be confirmed remarkably on the distribution of the time when the fibrils appeared, and there was a significant difference in the variance of the two distributions (Figure 7b) [93]. The bulk assay with gel filtration also indicated that the HMW incubation temporarily produced the LMW fraction [93]. There was no difference in the distribution of the fibril types between the incubation of the LMW and the HMW Aβ42 (Figure 7c) [93]. No significant difference was observed in the size of the fibril seeds at the time of appearance (Figure 7d) [93]. These results suggest that the on-pathway for fibril formation was common to both the LMW and HMW Aβ42 incubations, and that the HMW took time to form fibril nuclei [93]. As shown in Section 4.1, the HMW contained aggregates that dissociated according to the equilibrium shift after the addition of 0.1 M sodium chloride. Therefore, the HMW was located in the off-pathway and dissociated to generate the LMW; then, the fibrils were formed along the on-pathway [93]. This HS-AFM observation correlated to a recent kinetic study reported by Knowles et al., which calculated a fast Aβ42 oligomer formation (approximately 8 <sup>×</sup> 10−<sup>7</sup> S−1) and slow oligomer dissociation (approximately 9 <sup>×</sup> 10−<sup>5</sup> S<sup>−</sup>1) [136]. Thus, the delay in HMW dissociation to form a seed competent LMW is connected to the delay in fibril elongation.

*Int. J. Mol. Sci.* **2020**, *21*, 4287

**Figure 7.** Timing of the fibril appearance in the incubation of LMW and HWW Aβ42. (**a**) Time courses of cumulative number of fibrils in the observation area from LMW (blue lines) and HMW (red lines) Aβ42 incubation. Different lines correspond to different experiments. (**b**–**d**) The distribution of times at which individual fibril seeds appeared (**b**), the fibril type (**c**), and the length of fibril seeds when they first appeared in the observation area (**d**). The open and closed bars in (**c**) correspond to LMW and HMW incubation, respectively. Reproduced from [93].

#### **5. HS-AFM Observation of Early Aggregation Stages of A**β

Although highly challenging, there is considerable interest in probing the early events of amyloid aggregation to trace the formation of toxic oligomeric intermediates. While several biophysical approaches, including a combination of NMR techniques, have been employed to monitor these events at high resolution [48], they are limited by many factors, including the sample size, sensitivity, and timescale for measurements. On the other hand, HF-AFM is well suited for this purpose and can complement well with other studies by providing high-throughput measurements in real time.

Banerjee et al. characterized the structural features of low-order Aβ42 aggregates, measuring the spatial size of monomers through decamers and observing intramolecular structural dynamics of trimers, pentamers, and heptamers [109]. They prepared stable oligomers up to decamers by photochemical cross-linking and measured the sizes [109]. They found that the oligomeric order/size (volume) relationship showed two different proportionalities with the boundary at the tetramer, suggesting that at least two types of monomer packing patterns exist in their assembly [109]. Using HS-AFM, they observed intramolecular structural dynamics of the representative oligomers (trimer, pentamer, and heptamer) at subsecond temporal resolution [109]. The trimers sustained single blobs with almost constant width and length, while the pentamers and heptamers showed transition between a single compact globular shape and the extended multi-lobe structure (Figure 8). The pentamer had the two blobs with different sizes to each other in the extended state (Figure 8b). The heptamer can be extended to three lobes with different size, in addition to the double-lobe state (Figure 8c). These results suggest that Aβ42 oligomers are the lobe-linked structures, in which each lobe is composed of two or three peptides and that each oligomeric state can transiently extend and shrink between a compact globular shape and the multi-lobe structure [109].

Feng et al. identified that Aβ42 at the early aggregation stage can be classified into four structural types and characterized the kinetic interactions between those aggregates using HS-AFM [113]. They observed Aβ42 aggregates shortly after the dissolution of 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP)-treated Aβ42 in phosphate-buffered saline. Their statistical analyses based on the three dimensional size measurements indicated four classes of Aβ42 aggregates: Aβ15-20 nm with 15–20 nm (length and width) and height of 2.8 nm; Aβ36 nm of a bilobed structure with 34 nm length, 17 nm width, and 2.8 nm height; and AβAgg with 34–36 nm (length and width) and height of 9.2 nm; disordered chain-like structure [113]. They also observed binding/dissociating interactions between the different type aggregates. The on/off interactions of Aβ15–20 nm–Aβ36 nm and Aβ15–20 nm–Aβ15–20 nm were respectively characterized to be a single step process [113] (Figure 9b,c,e,f,h,i). The Aβ15–20 nm–Aβ36 nm binding showed slightly higher affinity compared with the Aβ15–20 nm–Aβ15–20 nm interactions [113]

(Figure 9h,i). In contrast, the binding time for Aβ15–20 nm–AβAgg was distributed randomly (Figure 9a,d,g), which suggests that the Aβ15–20 nm–AβAgg interaction was not a single step but accepted various binding patterns [113]. Inevitably, this "permissive" interaction may produce various types of oligomers with different toxicities [113].

**Figure 8.** *Cont*.

**Figure 8.** HS-AFM images of Aβ42 oligomers. Time courses of width and length obtained from HS-AFM images at the indicated times for trimers (**a**), pentamers (**b**), and heptamers (**c**). Reprinted with permission from [109]. Copyright (2017) American Chemical Society.

**Figure 9.** HS-AFM imaging and kinetic analyses of interaction between Aβ oligomers. Successive HS-AFM images (**a**–**c**), representative time courses of binding/dissociation (**d**–**f**), and distributions of the bound state (**g**–**i**) for the interaction of Aβ15–20 nm–AβAgg (**a**,**d**,**g**), Aβ15–20 nm–Aβ36 nm (**b**,**e**,**h**), and Aβ15–20 nm–Aβ15–20 nm. Reprint with permission [113]: Copyright (2019), Elsevier.

## **6. HS-AFM Observation of Interaction between Amyloidogenic Proteins and Other Chemical Compounds**

HS-AFM can also identify the interactions between amyloidogenic proteins and other chemical compounds, such as lipids and the potential anti-amyloidogenic inhibitor target(s) in the amyloid aggregation pathway by comparing observations in the presence and absence of the compounds.

#### *6.1. Interaction between a Toxic A*β *Oligomer and Lipid Bilayer*

Membrane–amyloidogenic protein interactions have been thought to lead specific aggregate conformers and play an important role in neurotoxicity as described in the Introduction. Aβ aggregation on a membrane has been known to depend on the physicochemical properties of the membrane (lipid composition, gel/liquid phases, phase separation, charge on head group, and oxidation level), which influences the toxicity of Aβ [137–143]. HS-AFM can visualize the effect of the lipid–amyloid interaction on the lipid membrane and characterize the interaction in real time.

Ewald et al. demonstrated the importance of the lipid composition on the interaction with a toxic Aβ oligomer, showing HS-AFM movies of lipid composition-dependent membrane disruption [111]. They prepared the toxic Aβ42 oG37C oligomers in which the 37th residue was changed from glycine to cysteine and added them to the planar lipid bilayer on the stage, changing the lipid composition in the mixture of sphingomyelin (SM)/1-palmitoyl-2-oleoylphosphatidylcholine (POPC)/cholesterol (Chol)/GM1-ganglioside (GM1). They visualized the GM1-dependent oligomer anchoring and the requirement of GM1/Chol coexistence at an appropriate ratio for membrane solubilization (Figure 10) [111].

#### *6.2. Aggregation Inhibition by Natural Phenolic Compounds*

Some natural phenolic compounds in foods have been known to inhibit amyloid aggregation [144–153]. The binding of the grape extract polyphenol myricetin to monomeric Aβ42 following the inhibition of amyloid aggregation was identified [152]. We investigated how myricetin altered the structural dynamics of Aβ42 amyloid fibril formation using HS-AFM. The observation was initiated by administering 2.5 μM Aβ42 (19:1 = LMW:seeds) and 10 μM myricetin to the HS-AFM sample chamber [93]. Some seeds appeared on the stage and they barely extended (Figure 11a) [93]. Then, the solution in the sample chamber was replaced with fresh 2.5 μM Aβ42 containing neither seed nor myricetin [93]. Immediately after the exchange, the fibrils started to grow, and the fibril amount also increased (Figure 11b) [93]. This result indicated that myricetin reversibly inhibited the fibril elongation reaction [93]. Further, the HS-AFM results suggested that the myricetin–monomer Aβ42 complex perturbed the dynamic equilibrium between the monomer and LMW that restricted the recruitment of LMW into the fibril ends. This could be by either the unavailability of seed competent Aβ42 species to recruit into the fibril ends or the reversible binding of myricetin to the fibril ends or both.

**Figure 10.** HS-AFM images of sphingomyelin (SM)/1-palmitoyl-2-oleoylphosphatidylcholine (POPC)/cholesterol (Chol)/GM1-ganglioside (GM1) membrane disruption after the addition of Aβ42 oG37C oligomers (**a**–**e**). The membrane with holes was removed from its edge. Reproduced from [111] with permission from The Royal Society of Chemistry.

**Figure 11.** HS-AFM observation of Aβ42 fibril growth in the presence/absence of myricetin. (**a**) HS-AFM images of LMW Aβ42 incubation with fibril seeds and myricetin at the indicated time after the addition of 0.1 M NaCl. Scale bar, 300 nm. (**b**) HS-AFM images of the same observation area after the replacement of the solution to fresh LMW Aβ42. (**c**) Schematic views of HS-AFM observation condition (i) and (ii) correspond to (**a**) and (**b**) Reproduced from [93].

#### *6.3. Aggregation Inhibition by Synthetic Polymers*

Amylin (also known as islet amyloid polypeptide protein (IAPP)) is a 37-residue peptide that is produced and co-secreted with insulin from pancreatic β cells. The amylin amyloid aggregates deposit on the β cells in the type II diabetes. Amylin shares some common characteristics with Aβ, such as folding into similar β-sheet structures [154], binding to amylin-3 receptor [155], and being digested by insulin-degrading enzyme [156]. Amylin crosses the blood–brain barrier (BBB) [157–159], and its aggregate deposition is found in the brains of type II diabetes patients with AD [121]. The mechanisms underlying the pathological [122,123] and suppressive [117–120] effects of amylin to AD remain controversial [160].

We observed amylin aggregation and fibril formation in the presence or absence of a polymethacrylate-derived copolymer (PMAQA) that has been used in various biological research fields, including lipid–nanodisc formation, Aβ–nanodisc interaction, the improvement of drug delivery, and the bioavailability on microencapsulation [41,161–163]. In this observation, amylin fibril seeds were immobilized on the stage beforehand, and then amylin monomers were added to the sample chamber alone or together with PMAQA [116]. In the absence of PMAQA, the original fibril seeds grew as observed in Aβ42, and some of the newly created fibrils in the chamber bound to and extended on the stage (Figure 12a,b) [116]. In the presence of PMAQA, the original fibril did not elongate, and de novo fibrils did not appear (Figure 12c,d) [116]. Unlike the HS-AFM observations for Aβ42 that showed fiber polymorphism (Figure 2), amylin showed a majority of straight fibers (Figure 12). NMR analysis of the PMAQA–amylin complex indicated that PMAQA bound to the amyloid core domain (NFGAIL) of amylin [116]. These results suggested that the binding of PMAQA to amylin monomer inhibited nucleation and self-replicative fibril elongation [116]. In this way, the consecutive imaging of fibril seeds adhered on the stage in advance and in the same observation area after the addition of

monomers could not only observe fibril elongation immediately, but also distinguished between the original fibrils and the de novo fibrils by identifying the appearance time and place of the fibrils.

**Figure 12.** HS-AFM observation of amylin fibril growth. (**a**,**c**) HS-AFM images of the amylin seeding reaction in the absence (**a**) or presence (**c**) of polymethacrylate-derived copolymer (PMAQA). The de novo nucleated fibrils are highlighted by yellow circles in (**a**). The growth extents of fibrils are indicated by the white arrows. (**b**,**d**) Kymographs of individual amylin fibrils highlighted by purple boxes in (**a**) and (**c**) in the absence (**b**) or presence (**d**) of PMAQA. The experimental setups are shown in (**e**). Reprint with permission [116]: Copyright (2019), the Royal Society of Chemistry.

The effects of styrene–maleic acid copolymers varying with charge (SMAEA/SMAQA) on amylin were investigated using HS-AFM [164]. The results from this study identified morphologically distinct amylin species. HS-AFM showed the cationic SMAQA polymer [165] interaction with amylin generates de novo spherical globulomers that are incompetent to grow in size or recruit to the fibril ends to proceed with the seeding reaction. In contrast, the anionic SMAEA polymer [166] accelerated amylin fibrillation and generated de novo spherical globulomers that grew in size and proceeded with the seeding reaction. These observations indicate that SMAQA and SMAEA acted as an inhibitor and promotor for amylin aggregation, respectively.

#### *6.4. Aggregation Inhibition by Heterologous Aggregation*

Although the heterologous aggregation of amyloidogenic proteins has been found in vivo, the molecular mechanism has been still unclear. In addition, designing peptides that lead to coaggregation with amyloidogenic proteins can potentially become the candidates for the therapeutic drugs.

Kakinen et al. observed the structural dynamics of homologous fibril growth of full-length amylin and the heterologous assembly of full-length amylin/its shorter component, 8–20 or 19–29 S20G, using HS-AFM [108]. The 8–20 and 19–29 S20G are the peptides from the 8th to the 20th and from the 19th to the 29th with replacement at the 20th residue of serine with glycine. Both of the two regions build up a cross β structure in amylin fibrils [108]. Using HS-AFM, they found and characterized the inhibition of amylin fibril growth in coaggregation with the peptides [108]. The fibril morphology of pure full-length amylin differed from that of coaggregation (Figure 13a–g) [108]. The fibril thickness of homologous full-length amylin was widely distributed from 9 to 20 nm, while that of the coaggregation fibrils showed a narrower distribution (Figure 13d) [108]. The authors interpreted that the mature fibrils formed at the pure full-length amylin on the HS-AFM stage, which was reflected in the wide distribution of fibril thickness (Figure 13d) [108].

Compared with the homologous aggregation, the coaggregation increased the surface roughness, which resulted from the higher production of small aggregates (Figure 13a–c,e–g). Similar to our Aβ42 study [93], kymographs showed that both the homologous and heterologous fibril growth of amylin followed the stepwise and polarized manner at the fast and slow ends (Figure 13h–j). The kymograph analysis also indicated that the coaggregation reduced the apparent growth speed (Figure 13k) due to decreases in the step speed (Figure 13l) and increases in the pause time (Figure 13m) [108]. These results were interpreted as the incorrect docking of the shorter peptides that should be removed or converted to the correct docking, which reduced the fibril elongation [108].

In addition, the authors focused on the relationship between step size and step time. This proportionality was much higher in the self-elongation of full-length amylin than in the coaggregation (Figure 13n) [108]. This analysis suggested that the step speed was kept at a constant value in the self-assembly of full-length amylin, while, for the coaggregation, the step speed was widely distributed [108]. They also showed significant recovery effects of coaggregation with 19–29 S20G on the survival rate, hatching rate, and phenotypic normality, using an in vivo model with zebrafish embryos [108].

**Figure 13.** HS-AFM observation of homologous and heterologous amylin fibril growth. (**a**–**c**) HS-AFM images of self-assembly of full-length amylin (**a**), coaggregations with 19–29 S20G (**b**) and 8–20 (**c**). (**d**) Distribution of fibril thickness. (**e**–**g**) Cross-sections at the dashed lines in (**a**–**c**). (**h**–**j**) Representative kymographs of homologous full-length amylin fibrils (**h**), heterologous fibrils with full-length amylin/19–29 S20G (**i**), and with full-length amylin/8–20 (**j**). (**k**) Apparent fibril growth speed. (**l**) Pause-free elongation speed. (**m**) Distribution of pause time. (**n**) Relationship between the step size and step time. Error bars correspond to the mean ± S.D. Asterisks represent statistically significant differences between the sample mean and control mean (ANOVA; \* *p* ≤ 0.05; \*\* *p* ≤ 0.01; \*\*\* *p* ≤ 0.005; \*\*\*\* *p* ≤ 0.001). Reprinted with permission from [108]. Copyright (2019) American Chemical Society.

#### **7. HS-AFM Observation of Other Amyloidogenic Proteins**

High-speed AFM reveals the structural dynamics of not only Aβ and amylin, but also other amyloid protein aggregations. Milhiet et al. showed protofilament elongtation and its stacking into polymorphic mature fibrils of lithostathine, which is overexpressed in the pre-clinical stage of AD [115]. Zhang et al. studied the structural dynamics of α-synuclein monomers and dimers using HS-AFM [110]. The monomer showed a transition between a spherical structure and a protruding tail-like structure and a structure completely extended, similar to a string [110]. The dimers were less flexible and basically maintained a dumbbell structure in which two spherical structures were connected [110]. Konno and Watanabe-Nakayama et al. observed yeast prion Sup35 monomer, oligomer, and fiber elongation [114]. HS-AFM revealed the structural dynamics of the intrinsically disordered (IDR) and partially folded regions of the Sup35 monomer, differences in the core structure and in the IDR

between Sup35 oligomers and fibrils, the stepwise growth of oligomers with distinct core size, and the continuous unidirectional elongation of Sup35 fibrils [114].

#### **8. Optimization of HS-AFM Observation of Amyloid Aggregation**

HS-AFM observation depended on the sample preparation, surrounding buffer composition, and imaging parameters. These conditions should be optimized so that the structure and dynamics on the HS-AFM stage are consistent with the conventional structural and dynamic analyses. In this section, we discuss the conditions for HS-AFM observation of amyloid aggregation.

#### *8.1. Sample Preparation and Control of Aggregation Initiation*

The sample preparation procedure should be optimized according to the aggregation process to be observed by HS-AFM. As shown in Sections 2 and 4, structural dynamics differ depending on the preparation of Aβ42 (LMW and HMW). In addition, the operational efficiency at the start of HS-AFM observation and the reproducibility of the results should be considered. The sample should be stored under conditions in which the aggregation does not progress, and the aggregation reaction needs to proceed during HS-AFM observation. In our studies, the size fractionated Aβ42 samples were stored in low ionic strength 10 mM sodium phosphate, at pH 7.4 [93,150], and then the aggregation was initiated by the addition of 0.1 M sodium chloride or potassium chloride [93].

#### *8.2. Sample Density*

Since the size of the HS-AFM observation field is limited as described in Section 8.4, a low density of sample molecules makes it difficult to find the molecules to be observed, while a high density means that the molecules are in contact with other molecules and have constrained structural dynamics; both cases should be avoided. For example, when the growing end of an observed fibril encounters another fibril, the growth stops there, and no further elongation occurs (Figure 1b). For the statistical analysis described in Sections 2.1 and 4.2, a sufficient number of aggregates are required within the observation area. Considering these conditions, the optimization of sample concentration is important. In our study, we set the concentration of Aβ42 to 2.5 μM in terms of monomers as in [93].

#### *8.3. Stage Materials and Sample Solution*

HS-AFM can only visualize the molecules at the interface between the solid surface of the stage and the liquid phase of the sample solution. HS-AFM cannot image molecules that are weakly bound to the stage and that are free in the sample solution. In addition, conditions in which molecules bind strongly to the stage could affect the structural dynamics and thus should be avoided. Therefore, we optimized a sample molecule binding condition in which the molecules stay at their positions during the time required to acquire their images in one frame, and where the structural dynamics are not affected (or the effect on the structural dynamics can be estimated). The binding force to the stage was determined by the chemical properties of the sample molecules (isoelectric point (pI), hydrophobicity, etc.), stage material, and composition of the sample solution. Thus, depending on the properties of the target molecule under investigation, one can optimize the chemical property of the mica surface (e.g., bare mica or 3-aminopropyltriethoxysilane (APTES) modified mica) and/or solution pH and ionic strength to obtain the desirable structural and dynamic information.

An atomically flat surface of freshly cleaved mica ([KAl2(OH)2AlSi3O10]) was used as a first candidate for the HS-AFM stage. The surface was negatively charged. The pH and ionic strength of the solution and the pI of the sample molecules determined the electrostatic interaction with the mica surface. As the pI of Aβ42 was 5.31 (estimated by the ProtParam [167] in Expasy), the net charge was negative in a neutral pH solution, and electrostatic repulsion was expected to act between the peptide surface and mica. In fact, we observed Aβ42 to be less adsorbed to mica in a low ionic strength buffer of 10 mM sodium phosphate, pH 7.4.

Electrostatic interactions between mica and sample molecules can be controlled by the addition of salts, as observed in ion exchange chromatography. Aβ42 aggregates bound to the mica surface immediately after the addition of 0.1 M sodium chloride to its low ionic strength buffer solution (10 mM sodium phosphate, pH 7.4), suggesting that the electrostatic repulsion acting between the Aβ42 and mica surface was canceled by the addition of salt and the hydrophobic interaction bound Aβ42 to the mica surface. The effect of salt on the electrostatic interaction between sample molecules and the mica surface varied depending on the type of cation, even at the same concentration. Potassium ions had a greater canceling effect of this electrostatic interaction than did sodium ions [168,169]. We used this difference in potassium and sodium ions to change the interaction between Aβ42 and the mica surface to characterize the structure switch of Aβ42 fibril elongation, as described in Section 2.1.

Chemical modification of the mica surface or other materials can be used to modulate the interaction between sample molecules and the stage surface. Modification of the mica surface with 3-aminopropyltriethoxysilane (APTES) made the surface positively charged [170]. In addition, the amino group of APTES can be used for the covalent immobilization of sample molecules with glutaraldehyde [170]. Highly ordered pyrolytic graphite (HOPG) can be used for the immobilization of sample molecules with hydrophobic interactions. HOPG was used for the immobilization of Aβ25-35 fibrils [126].

The effect that the sample-stage interaction had on the structural dynamics of sample molecules needed to be verified. To examine this effect, we confirmed the consistency between the time courses of the ThT assay in the in vitro aggregation reaction and the total amount of aggregate in the HS-AFM observation area. The trends of the total aggregate amount and the ThT fluorescence intensity were consistent with each other (Figure 14a,b) [93]. They rapidly increased in the HMW Aβ42 incubation and gradually increased at a low level in the LMW Aβ42 incubation (Figure 14a,b) [94]. The aggregate structures were also consistent between the HS-AFM image and the transmission electron microscope (TEM) image observed at the same time points after the initiation of aggregation (Figures 1b, 6a and 14c,d). For our HS-AFM observation of the amylin aggregation, we used a low ionic strength solution and confirmed no significant difference in observation between bare and APTES mica, indicating that the electrostatic interaction between the sample and the surface did not affect the structural dynamics of amylin in our observation [116].

**Figure 14.** Relationship between HS-AFM observation, thioflavin T (ThT) assay, and transmission electron microscopy (TEM) of LMW and HMW Aβ42 incubation. The time evolution of the total Aβ42 aggregate density in HS-AFM observation (**a**); time course of ThT fluorescence intensity (**b**); TEM images of LMW (**c**; blue lines in (**a**,**b**)) and HMW ((**d**); red lines in (**a**,**b**)) Aβ42 incubations. Bars, 100 nm. Reproduced from [93].

#### *8.4. Size of Scanning Area and Time for Image Acquisition*

The observation area and imaging speed are in an inversely proportional relationship. For the statistical analysis of structural dynamics of amyloid aggregates, the size of the observation area is needed prior to the imaging speed. In this case, the typical scale of the observation field and the typical imaging speed are micrometers and several seconds to ten seconds per frame, respectively. For the imaging of faster structural dynamics of individual aggregates, the typical imaging speed is subseconds per frame, which requires a reduction in the observation area. The time 'T', also referred as the scanning rate, for the acquisition of one frame in the observation area (W nm in width × N lines in the y-direction) is expressed as follows [170]:

$$\mathbf{T} = \pi \mathbf{W} \mathbf{N} / (2\lambda \mathbf{f}\_{\mathbf{B}} \theta\_{\mathbf{m}}) \tag{3}$$

where λ is the periodicity of the sample surface with the sinusoidal shape, fB is the feedback bandwidth, and θ<sup>m</sup> is the maximum allowable phase delay of feedback control. λ is usually several nm, fB is about 100 kHz at maximum, and θ<sup>m</sup> is ≈20◦ or less for fragile samples and ≈45◦ or more for stiff samples. In addition to Equation (3), the number of frames captured in the same observation field should be considered. The tapping HS-AFM probe applies a mechanical force to the sample molecules. Thus, the larger the number of frames, the higher the probability that the sample will be broken. In addition to the imaging speed and applied force, the number of frames should be reduced for the analysis of increasing numbers of amyloid fibrils because the fragmented amyloid fibrils individually serve as seeds and then grow.

#### **9. Conclusions**

Different Aβ fibril structures have been found in different AD patients, which evokes the relationship between the fibril formation and AD progression. The structural dynamics of amyloid aggregation need to be elucidated for diagnosis and drug discovery. In previous studies, the structural dynamics were characterized by separately studying the structure and dynamics. However, the conventional structure and dynamics analyses lack information from the other missing component. Thus, the information that could not be gathered from those studies has remained unknown. The development of HS-AFM allowed video recording with nanometer spatial resolution, which has enabled the simultaneous analysis of structure and dynamics. We were able to analyze the structural dynamics of individual amyloid aggregates, including fibrils, even when different types of aggregates coexisted.

Recent studies have shown that amyloid proteins formed aggregates with diverse structures under physiological conditions that differed from those in vitro [171]. Under physiological conditions, amyloidogenic proteins underwent interactions with biological membranes, metal ions, other amyloidogenic proteins, and variants with different amino acid sequences causing heterologous aggregation. When reproducing the various physiological conditions by immobilizing a planar membrane on the stage and/or adding metal ions or variants of amyloidogenic proteins to the sample chamber, HS-AFM was used to visualize the structural dynamics in the aggregation processes. In addition, comparison with the observations in the presence of designed inhibitors may contribute to therapeutic development to characterize the desired drug targeting specific toxic aggregates.

HS-AFM observation has some unique characteristics: the visualization was limited to the structural dynamics that occurred in the solid–liquid interface of the stage and the sample solution; the imaging speed and size of the observation area were restricted by their inverse proportion relationship; and the statistical analysis and the macroscopic trend of the HS-AFM observations must be confirmed to be consistent with the conventional structure and dynamics analyses. In this way, HS-AFM did not only link the previous structure and dynamics analyses but did identify the structural dynamics that could not be elucidated using conventional methods.

**Funding:** This work was supported by Kanazawa University CHOZEN project. Resources for the amylin study were supported by NIH (AG048934 to A.R.).

**Acknowledgments:** We thank Hiroki Konno for his help and Noriyuki Kodera and Toshio Ando for their development of HS-AFM.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **Abbreviations**

