Amyloids are fibrous protein aggregates defined by shared specific structural features. Abnormal accumulation of amyloid in organs leads to amyloidosis, which results in various neurodegenerative diseases. Atomic force microscopy (AFM) has proven to be an excellent tool investigating amyloids; it has been extensively utilized to characterize its morphology, assembly process, and mechanical properties. This review summarizes studies which applied AFM to detect the inter- and intramolecular interactions of amyloid fibrils and classified the influencing factors of amyloid’s nanomechanics in detail. The characteristics of amyloid fibrils driven by inter- and intramolecular interactions, including various morphologies of amyloid fibrils, self-assembly process, and the aggregating pathway, are described. Successful examples where AFM provided abundant information about inter- and intramolecular interactions of amyloid fibrils in different environments are presented. Direct force measurement of intra- or intermolecular interactions utilizing an AFM-based tool, single-molecular force spectroscopy (SMFS), is introduced. Some mechanical information such as elasticity, adhesiveness, and strength was obtained by stretching amyloid fibrils. This review helps researchers in understanding the mechanism of amyloidogenesis and exploring the properties of amyloid using AFM techniques.

1. Introduction

Improper aggregation of polypeptide fragments may result in various neurological disorder diseases [1], such as Alzheimer’s disease (Aβ aggregation) [25], Parkinson’s disease [6], Huntington’s disease (Huntington aggregation) [7, 8], prion disease (PrP aggregation) [9], and amyotrophic lateral sclerosis (ALS) [10]. Amyloid aggregations are also found in type II diabetes (islet amyloid polypeptide) [1114] and dialysis related amyloidosis (β-2 microglobulin aggregation) [6]. Recently, more and more studies have suggested that these diseases are related to the aggregations formed by amyloids sharing specific structural traits. Single soluble amyloid proteins start to interact with each other, and these intermolecular interactions finally assemble the soluble amyloid into various insoluble forms. In addition, a great variety of heterogeneous morphologies detected in self-assembly processes indicate different assembly pathways of amyloid fibrils [1517]. As a whole, their assembly pathways can be simply described as soluble protein nucleation fibrillar elongation/lateral aggregation mature network [18]. Though multiple amyloids have been widely explored in recent years, their pathogenic mechanism has not been elucidated clearly.

Atomic force microscopy (AFM) is an excellent tool which has been used extensively to study the fibrillar ultrastructures. AFM enables us to clearly visualize individual biological macromolecules at the nanometer scale [1923]. Time-lapse AFM imaging [24] has been successfully adopted to monitor the growth of individual peptide fibrils and to characterize the influence of the chemical environment on amyloid aggregation [25]. AFM-based single-molecule force spectroscopy (SMFS) [26] has made force measurement at the single-molecule level with pico-Newton (pN) force resolution possible. This technique enables researchers to analyze inter- and intramolecular interactions [27, 28]. The obtained mechanical fingerprint [29] of amyloid fibrils has proven that SMFS is an efficient tool to explore the mechanisms of amyloid assembly process, the differences of various amyloids, and the mechanisms of interactions with chemicals and chaperones.

We review AFM techniques applied to characterize and understand the assembly process of amyloid fibrils involved in pathogenic disorders. We summarized AFM studies of amyloid fibrillogenesis focusing on the morphology, kinetics, and models of amyloid self-assembly. The investigation of the inter- or intramolecular interaction of amyloid fibrils using AFM-based SMFS was reviewed to explore the assembly mechanism and mechanical properties of various shaped fibrils, such as globular oligomers, protofibrils, mature intertwined fibrils, and network structures.

2. Principle of AFM-Based SMFS

AFM is a type of Scanning Probe Microscope (SPM) [65] with high-resolution, and now it has become one of the foremost tools in imaging, measuring, and manipulating matter at nanoscale [6669]. AFM allows imaging both in ambient and in liquid environments which is of great importance for biological molecules [7073]. AFM-based SMFS stands out among various single-molecule techniques [7477] because of its high detection rate, easy operation, and wide application in measuring weak inter- or intramolecule interactions [7880]. Through SMFS technique, multiple properties, such as elasticity and viscosity [81], can be analyzed in detail. At the same time, SMFS is analytic technique applied not only to measure mechanical properties of various proteins but also to manipulate single-molecule at pico-Newton scale [78], for example, probing the helical structure, unfolding β-fold structure [64], and measuring intermolecular interactions.

Compared with simple topographic characterization, the AFM-based single-molecule force spectroscopy is more complex [26]. SMFS measurement is based on full knowledge of the force on the tip during approach and retraction. During the approach, the AFM tip moves slowly toward the sample surface and the cantilever is bent toward the sample surface when the tip-sample distance arrives at a certain force-distance, as it starts to feel long-range attractive forces like van der Waals forces (Figures 1(a), (1)-(2), and 1(b), A-B). With further movement, the cantilever starts to feel repulsive forces as well; it eventually recovers equilibrium. Further approach to the sample increases the repulsive force that will bend the cantilever away from the sample, and the tip finally stops at the predefined force value (Figure 1(a), (4)-(5)). The compressive stiffness or elasticity modulus of the molecule can be deduced from the force-distance curve between C and D from this approaching process. During retraction the probe slowly moves away from the substrate and fingerprint information representing the molecular mechanical properties is captured during the stretch-relaxation process. In the retraction force curve (Figure 1(c)), point E represents an adhesion force caused by nonspecific interaction between the AFM tip and substrate. Point F indicates the start of the phase when the molecule is pulled away from the substrate. The force increases until point G when the molecule ruptures from the AFM tip or substrate, and cantilever returns to the equilibrium position (point H). The nanomechanics such as the stretching or unzipping response of the molecule can be obtained from this pulling process.

The mechanical responses observed in SMFS can be divided into three categories: single nonlinear elastic curve, saw-tooth curve, and irregular mechanical event [64, 82]. The long single nonlinear elastic curve is characterized by a large peak which indicates the detachment of the sample from the tip. Some groups reported that peaks in the force plateau [82] were caused by simultaneous unzipping of the molecular strand. The saw-tooth force curve is composed of several peaks starting with irregular peaks and ending with a rupture force. Each peak represents an unfolding or sudden detachment event. The plateau force curve has a long uniform plateau and an abrupt force drop belongs to the manipulating response of β-sheet [64]. An irregular mechanical event generally refers to complex multimolecular interactions [83] in which interconnected fibrils are stretched sequentially leading to the extension and breaking of loops or bonds in one pulling cycle. Among the mechanical responses, single nonlinear and orderly force patterns are useful in exploring mechanical properties. At the same time two prominent theoretical models, the Freely-Jointed Chain (FJC) model [84] and the Worm-Like Chain (WLC) model [85], have been developed to analyze these SMFS data.

3. Intermolecular Interaction

The assembly of amyloid peptides is a dynamic process. The pathway from soluble molecules to insoluble fibrils is driven by intermolecular interactions which usually result in the formation of heterogeneous shaped structures. Amyloid fibrils interact with each other and aggregate into larger fibrils and eventually transform into texture structures. According to previous reports, the textured structures are constituted by different mature fibrils generated through multiple pathways. Therefore, obtaining the morphologies of various structures at different stages is essential to comprehend their assembly process.

3.1. Self-Assembled Structures

Plenty of shapes of amyloid structures were reported in many previous studies, including globular, β-hairpin, β-sheet, disk-like, worm-like, rod-like, honeycomb, parallel, and braided structures [1]. The variety of these assembled structures is attributed to the intermolecular interactions including hydrogen bonds, electrostatic interaction, and hydrophobic interactions [32, 38, 8689]. The globular structures formed at the beginning of the incubation process are easily observed in AFM (Figure 2(a)). They consist of many monomers as well as disk-like structures [16, 39, 44, 58]. Branch-like [14] and parallel structures [38] (Figures 2(b) and 2(c)) are the intermediate products [90] composed of several oligomers. A mature fibrillar structure (Figure 2(d)) is a large fibril composed of two or more fibrils. Table 1 shows the dimensions of fibrillar structures obtained at different incubation stages, and the stages are represented by their typical structures.

McAllister et al. [87] found that the increase of protein-protein interaction usually resulted in morphological transformations, for example, β-sheet conformation with an elevated content. Gerber et al. [91] have reported disk-like structures that form stacks through interoligomer interactions. Sandal et al. [64] have studied β-like formation of α-Syn and found the relative abundance of the β-like structures significantly increased in different conditions promoting the aggregation of α-Syn, such as pathogenic A30P mutation and high ionic strength buffer. Sibley et al. [86] found that the interaction between insulin and porphyrin gave rise to circular, ring-like structures as well as fibrils. The possible reason to form various morphologies is related to the interactions between specific residues. Jansen et al. [38] found that the compact character or mature fibrillar structures might originate from the effort to minimize the exposure of hydrophobic residues. In order to explore the effects of specific residues, some functional residues were substituted during AFM-based measurements. Various mutations showed distinctive functions: some were prone to form amyloid fibrils while others formed spherical aggregates; some functioned as a β-sheet breaker while others were promoting overall-length aggregation [32, 9294]. For example, Aβ40 and were found to form small oligomers and thin fibrils, respectively [94]. However, A53T and A30P mutants of α-Syn were found to form spherical or annular protofibrillar structures [34]. Common morphologies were observed for some residues, for example, twisted fibrils derived from the Q24K mutant, and spherical aggregates and short fibrils derive from other mutants. E46K mutant displays a very distinctive smaller periodicity [36] compared with other mutants. We summarize the different shapes of amyloid with different dimensions in Table 1.

3.2. Assembly Processes

Time-lapse monitoring of the amyloid aggregation process is crucial to deepen the understanding of the amyloid aggregation mechanisms. The assembly of various amyloid fibrils can be followed by in situ time-lapse AFM images. Amyloid aggregation is commonly divided into two stages: nucleation stage and fibril growth stage [47].

In the nucleation stage, often called lag-phase, it is critical to understand the behavior of “seed-like” structures and intermediate prefibrillar structures, as these are the starting point of the overall self-assembly process [95, 96]. Fukuma revealed that the lag-phase was related to the increase of the mass concentration of elongated fibrils, and long incubating time was not an important factor during the nucleation stage [38, 97]. When studying on the process of Aβ aggregation, Harper et al. [98] found that the rate of oligomers was slower than that of fibrils and that fibrils rapidly aggregated once sufficient nucleated oligomers formed. Their results indicated that the elongation rate of individual amylin protofibrils was  nm/min. In line with the aggregation pathway, the stability of monomer and oligomer state was significantly lower than that of the following stages. It is reported that the inhibition of fibril formation could be realized by reducing the stability of protofibrils, by blocking protofibril-protofibril interaction or by shifting the protofibril-monomer equilibrium. Oligomers contain nonfibrillar β-structures, and their total amount remains almost constant from the second half of the nucleation phase to the end of the aggregation process [44].

In the fibril growth stage, growth rate and aggregation propensity of amyloid assembly are influenced by different sequences or specific residues of the peptide. The corresponding amyloid assembly has been investigated by substitution of residues. Various amyloid peptides and their mutants were studied, such as Aβ with mutant AβE22G, , or Aβ40ARC [59, 99, 100], α-synuclein with disease-related A30P, E46K, and A53T variants [34, 36, 93, 101, 102], mutant huntingtin (Htt) [103, 104], PAPBN1 N-WT with N-(+7)Ala mutant [105], and β2-microglobulin with its deamidated variant N17D [106]. Another example, rat amylin, although 84% residues are the same as in human amylin, cannot form amyloid fibrils [107]. The possible reason is that those residues, which differ from human amylin, influence the peptide assembly [59, 94]. In Aβ mutation (AβE22G), fibrillization process will be accelerated, while the abundance of nonfibrillar assemblies will be decreased. Conway et al. [101] reported that the fibrillation rate of specific mutant peptides or mutant mixtures was faster than that of WT peptide. Seed-induced fibrillation of N-WT of PAPBN1 was slower than that of N-(+7)Ala. Monitoring the solubilization kinetics, they found that the stability of N-WT and N-(+7)Ala fibrils was different. In another case [98], and formed two discrete morphologies, and aggregates grew faster than ones. However, the rate of Aβ amyloid aggregation in vitro was limited by the amount of available Aβ nuclei. Moreover, the amounts of aggregated and protofibrils obviously differed from each other. Marek et al. [92] suggested that the difference between the amount of aggregated Aβ1-40 and Aβ1-42 protofibrils was caused by different residues affecting the aggregating rate of fibrillogenesis. In their study, the kinetics of amyloid assembly and the resulting morphology were influenced by the aromatic residues, which were important during the lag-phase in AFM measurements. Table 2 shows the assembly parameters of various amyloids under distinctive incubating environments. This overview suggests that experimental factors, such as buffer, pH, temperature, and concentration, are critical to the result of the fibrillation process.

3.3. Assembly Pathways

The aggregation process was reported to be associated with the pathology of the corresponding amyloid protofibrils. Numerous studies have been carried out to explore the aggregation pathway [13, 108]. For the mechanism of amyloid fibrillogenesis, several explanations have been established. It is suggested that the common noncovalent structure of proteins such as backbone hydrogen bonding and hydrophobic interaction [17] were the main forces driving the amyloid fibrils’ aggregation. In early studies, a mechanism of nucleated conformational conversion, so-called on-pathway, was applied to explain the amyloid aggregation. However, exceptions have been found. Therefore, an alternative off-pathway mechanism was proposed to explain fibrillogenesis [57, 63]. While more and more studies explore the mechanism of amyloid aggregation, models for various kinds of amyloid fibrils have been designed to explain the amyloidogenesis formation. Here, we propose a model (Figure 3) based on various previous studies [1517, 34, 38, 46, 53, 56, 62, 63, 109113] to elucidate the mechanism of multipathway aggregation and describe it in detail in the following parts.

3.3.1. Nucleation and Elongation

AFM measurements revealed that the most favorable nucleation pathway contains a two-stage sequential conversion (Figure 3, steps 1 and 2), in which soluble monomers are aggregated into small annular and spheroidal mature oligomers [14, 114] and then these seeds grow by further addition of more mature monomers. Mature oligomers have accumulated more monomers but still show globular morphology. Oligomers still have spherical superstructure but already show characteristic amyloid folding [93]. [45, 47, 48, 115, 116], glucagon [56, 117], amylin [24, 107], and β-lactoglobulin [49, 50] have been observed to aggregate through the nucleation pathway. Fibril elongation (Figure 3, steps 4, 5, and 6) becomes the main process once a critical amount of oligomeric seeds has formed. In the elongation process, the addition of more monomers leads to a structural change into elongated prefibrillar intermediates, eventually resulting in the formation of protofibrils [45, 105]. Different assembly processes of amyloid were indicated in different color of lines in Figure 3.

3.3.2. Hierarchical Pathway

Hierarchical aggregation, which happens after nucleation and elongation, is characterized by two or more protofibrils intertwining through interoligomer or interfibril interactions. They form higher ordered fibrils and eventually helical structures. Many species, CP [17, 118], human prion protein (PrPSc) [16], [119], Ig light-chain [62], transthyretin peptide () [120], and β-lactoglobulin [51], were found to aggregate adopting a hierarchical pathway. Small or large oligomers undergo elongation and form heterogeneous structures, such as branch-like structures, annular-shaped oligomers, braided structures, and hairpin-like structures [17, 45, 46, 55]. Sbrana et al. [17] reported that branched structures were the disordered assembly of protruding segments. They also found that early annular-shaped oligomers seem to function as fundamental bricks in the hierarchical aggregation process [17]. The braided structure [62] consisting of winding protofibrils is usually observed in amyloid fibrillogenesis as well. In the self-assembly experiment of Aβ42 peptides [48], intermediate-like protofibrils were found to join the helical structure formation. Generally speaking, these heterogeneous morphologies and twisting periodicity indicated a complex hierarchical amyloid assembly process.

3.3.3. Lateral Aggregation

Increasing evidence suggests the existence of alternative pathways [38] in amyloid fibrillogenesis. One prominent example is lateral aggregation; it usually follows the elongation phase. Ceratoplatanin (CP) [17], PrP [109], glucagon [56, 95], insulin [38, 57], [46, 47], and β2-microglobulin [53, 121] were found to aggregate laterally. In this pathway several protofibrils associate parallelly to form a ribbon that wraps around into a fibril (Figure 3, steps 10 and 12, type 3 structure). It was reported that fibrillar bundles formed loose tangles eventually leading to the formation of mature fibrils [57]. Fibrils containing laterally associated filaments were found to show a right-handed twist at one point [47]. A similar aggregation pathway was also found in the strand-swapping peptide 1 (SSP1). Nagarkar et al. [55] reported the lateral self-assembly of SSP1 dimers via H-bond interaction along the fibril’s long axis. Kad et al. [106] reported that four protofibrils associated laterally wound into a twisted-ribbon shape with a clear periodicity, but there was no suggestion that lateral aggregation of smaller species was detected [49].

3.3.4. Multipathway

Multipathway is the combination of all pathways mentioned above: monomers conformationally change and merge into oligomers (Figure 3, steps 1 and 2); then oligomers longitudinally aggregate leading to protofibrils (Figure 3, steps 4, 5, and 7). Finally protofibrils laterally aggregate into protofibrils (Figure 3, steps 6, 10, and 11) [122]. Homogeneous protofibrils undergo elongation to form higher ordered mature fibrils (Figure 3, steps 13, 15, and 16) and finally lead to complex blocks. Hierarchical and lateral-aggregating structures were frequently observed in various kinds of amyloid fibrils [1517, 36, 38, 45, 48, 53, 56, 57, 62, 95, 109, 111113, 123].

In the off-pathway assembly, soluble monomers or oligomers directly construct fibrils [63] without the “seed-like” aggregation (Figure 3, step 3). Natalello et al. [16] reported that the linear aggregates formed by oligomers aligning which suggested an off-pathway assembly. The main differences between on- and off-pathway oligomers are mainly their sizes and shapes. So, it is critical to clarify whether the aggregation is based on a nucleation phase and seeds or the formation of an active small oligomer.

On-pathway aggregation is characterized by the appearance of homogeneous nuclei, followed by elongation. At the same time, three types of fibrils (Figure 3) were found during the later stage of the aggregation process. These different types represent distinctive structures: type 1 is formed by two twined protofibrils; type 2 is formed by three twined protofibrils; and type 3 is formed by several parallel protofibrils laterally associated together. Several type 1 fibrils rearrange into intertwined style fibrils occasionally. Based on associated segments forming larger structures, Segers-Nolten et al. [36] proposed a segment pathway, indicating a multipathway assembly for α-synuclein. Jansen et al. [38] revealed that insulin amyloidogenesis in vitro involved a multipathway assembling scheme, in which native dimers were formed by either hierarchical intertwining or lateral interaction. A similar observation was made by Mauro et al. [57]. The size and shape of oligomers were measured to identify different distinctive pathways. However, it could not be distinguished [55] whether the hierarchical or parallel fibrils were lacking structurally different nucleating centers. Although various models have been proposed, the detailed mechanism needs further exploration such as amyloids’ aggregation and inter- or intramolecule interactions affected by constituent peptides or chemical chaperones.

3.4. Influencing Factors of Assembly

High-resolution AFM has been used in many characterization studies, aimed at the morphology and assembly pathway of amyloid fibrils and the effects of chemicals and chaperones [14, 25, 73, 124126]. Concentration [36, 51, 96, 127129], substrate [130], temperature [38, 49, 50, 57, 131], pH value [54, 106, 121, 132, 133], ionic strength [96, 99, 129, 134, 135], and stirring time and addition of denaturing agents [24, 50, 82, 115, 119, 128, 136] are important factors affecting the formation of various aggregates. For example, different substrates can affect the orientation of amyloid fibrillogenesis [4]; solvent conditions play critical roles in amyloid aggregating propensity, rate, and structural formation. In order to decipher the molecular mechanisms and develop better strategies to modulate aggregation, it is imperative to learn the effects of environmental conditions on structure, molecular assembly process, activities, and growth kinetics. In this section, we will have a closer look on these experimental factors.

3.4.1. Concentration Effect

Many trials indicated that the concentration of amyloid peptides played a prominent role in amyloid aggregation. Although differences in concentration are correlated to the corresponding disease in vivo, their precise relation is not well-known. Previous work showed that the self-assembling rate of amyloid increased with the increasing of its solution concentration [137]. Segers-Nolten et al. [36] found that α-synuclein shows relatively normal function at low concentrations, but it is apt to transform into a pathogenic species at high concentrations. There are a large number of experiments looking into surface density and concentration of the incubation solution [36, 51, 96, 127129]. These experiments indicate that proteins form well-defined fibrils in low peptide concentrations with lower aggregation rates than that in a higher concentration. So, the amyloid fibril formation can be accelerated through increasing either surface density or the concentration in incubation solution [127]. In the same way, Pazzagli et al. [118] systemically studied the lag-phases in different concentrations and found that the transition time in higher concentration (1.3 Mm, lag-phase time being 6 hours) was sharply shortened comparing with that of the lower concentration (0.54 mM, lag-phase time being longer than ten days). The surface density of amyloid self-assembled fibrils can be adjusted by tuning the bulk concentration, and many groups showed that dense fiber-networks can be constructed starting with high peptide concentrations [96, 129]. However, insulin is an exception, as obvious structure change was observed for two enormously different concentrations [57].

3.4.2. Temperature Effect

Temperature, in general, can affect morphology, growth rate, stability, and activity of heterogeneous fibrils and eventually change the overall process of amyloid aggregation [38, 49, 50, 57, 131]. For example, various structures were observed upon increasing the temperature to 70°C, among them long straight rods, twisted-ribbon-like structures, rod bundles, and rope-like structures [38]. Increasing temperature can not only shorten the aggregation lag-phase [118] but also affect the height of the assembled fibrils [138]. In contrast, nucleation was inhibited at low temperatures. Pazzagli et al. [118] investigated the ordered aggregates of ceratoplatanin and found that lag-time decreased from 30 to 10 days when incubation temperature was increased from 37°C to 50°C. Palhano et al. [60] employed 4°C and 25°C to investigate the effect of temperature on the process of amyloid aggregation. Their AFM results showed that the aggregations were higher at 4°C than at 25°C. At the same time they revealed that amyloid fibril were, on average, shorter at 4°C than at 25°C. The reason for this phenomenon is that the activity of amyloid can be influenced by temperature [131]. Native Mβ activity remained stable up to 70°C, but its activity abruptly decreased at a temperature ranging from 70°C to 80°C. Bellezza also found that the main activity of adsorbed Mβ decreased abruptly between 30°C and 60°C, while the activity reduced slightly below 30°C or higher than 60°C. Mauro et al. [57] studied the temperature impact on insulin and found the assembly adopting double quenching experiments. Their results indicated that the double quenching allowed the growth of a few long fibrils.

3.4.3. pH Effect

An acidic environment is beneficial for the amyloid formation [35, 87, 106, 128], and there are plenty of studies modulating amyloid aggregation through varying pH values [54, 106, 121, 132, 133]. In these studies, researchers found that amyloid fibrils were not stable in either acidic or alkaline solution environments, which easily led to the conformational changes. Many investigations suggested that the nanostructure of various amyloid assemblies could be modified through adjusting pH value. Bortolini et al. [133] built different nanostructures of peptide with three different kinds of residues by tuning the pH value of the solution [133]. McAllister et al. [87] reported that decreasing pH value resulted in the prominent increase of the interaction among protein molecules of Aβ (1–40) peptide, α-synuclein, and lysozyme. This leads to a dramatic increase in aggregation rate at the proper pH value. For most peptides there are large differences in reaction speed and product morphology between acidic and alkaline conditions. Short fibrils or small globular aggregates were found at pH 2.0, and fibrillar structures were found at pH 2.7, but there was no fibril or large aggregate observed at pH 3.7. Hong et al. [54] studied two kinds of amyloid aggregates at pH values varying from 4 to 11. Hong et al. found that KAK16-IV formed globular assemblies in neutral pH environments, which changed into fibrils under alkaline conditions. Another mutant, EAK16-II, did not exhibit any apparent changes. Jenko et al. [139] established that Stefin B started to form fibrils at pH 5, whereas Stefin A needed to be acidified to a pH value of less than 2.5. Most tests showed that acidic environments were conducive to fibrils formation, but the transformation of Stefin B from protofibrils into mature fibrils was inhibited at acidic solution [128].

3.4.4. Solvent Effect

AFM experiments suggest that the incubating medium plays an important role in the assembly process [47, 4951, 97, 127, 140, 141]. Chaudhary et al. [97] reported that AcPHF6 could be organized into fibrillar structures when the sample peptide was dissolved in MeOH, TFE, or HFIP. Gosal et al. [49] found that the aggregation rate of β-lactoglobulin was correlated with solvents used in experiments. There are more fibrillar structures presented in TFE-water mixed solvent in contrast to other alcohols. Gelling propensity was related to solvents: methanol > ethanol > propanol > TFE. At pH 7, the tendency of β-lactoglobulin to form a gel was higher in propanol than that in ethanol, methanol, or TFE. The fibrillar aggregates formed in TFE-water mixtures; imaged with negative-staining EM these TFE-induced fibrils showed worm-like and granular structures [49]. Nichols et al. [43] found rapid assembly of amyloid-β peptide at a liquid/liquid interface which induced unstable β-sheet fibers. The association rate of in a two-phase system with chloroform was 1~2 orders of magnitude faster than that in the buffer alone. Daniela et al. [50] reported that β-lactoglobulin formed different amorphous aggregates in alcohols and TFE. The concentration of TFE also influenced the assembly process [128]. The aggregating rate of human Stefin B fibrils was accelerated in a solution containing alcohol, but, in contrast to other proteins, the lag-phase did not change TFE concentration.

3.4.5. Cations Effect

Metal ions such as Fe3+, Cu2+, K+, and Na+ significantly affect the process of amyloid aggregation and morphology [96, 99, 129, 134, 135]. Ions function as an inhibitor or accelerator to various amyloid species, and fibril shapes may be influenced by varying the metal ions concentration. Precisely how the ions effect the aggregation is still controversial. Ryu et al. [96] reported that the initial rate of amyloid fibrillation was accelerated by 6 times in the presence of Fe3+ ions, but ions might act as an inhibitor under other conditions. For example, high ion concentration inhibited amyloid aggregation of rat amylin [107]. In addition, disrupted adhesive nanofiber structures can be repaired by solutions containing divalent cations [127].

Apart from the effects on amyloid aggregation speed, the morphology is also influenced by cations [134], as they interfere with peptide-peptide interaction. For example, fibrillar structures tend to form at low Cu2+ concentration, but the amount of granular, amorphous aggregation increased rapidly at higher concentrations of Cu2+. Hong et al. [129] reported that the dimensions and surface tension of peptide nanostructures were influenced by the NaCl concentration in the solution. The orientation of amyloid aggregation on mica was affected by ions [99]; this was attributed to cooperative interaction of a positively charged peptide moiety binding to the mica lattice. They pointed out that binding to mica was sensitive to the presence of cations and suggested that the increase of NaCl or KCl concentration could reduce the binding strength between fibrils and mica surface. Further research indicated that fibrils binding to mica were more sensitive to K+ compared to Na+ ions.

3.4.6. Denaturing Additives Effect

Various additives are usually employed to modulate the behaviors of amyloid fibrils through accelerating, inhibiting aggregation, or disassembling [24, 50, 82, 115, 119, 128, 136]. Some reversible changes can be accomplished [127, 142] by varying the concentration of additives. Different additives have been used to study their effects on various amyloids, such as the effect of Zn, sulfated glycopolymers, C12C6C12Br2 micelles, Trimethylamine N oxide, and glycerol on Aβ peptide [82, 141143], TFE on human stefin B [135], anionic lipid phosphatidylserine (PS) and cholesterol on amylin [24], antibody scFv on α-synuclein, insulin, and β-amyloid [136, 142], chitotriose and NAG on HEWL [132], DTT and SDS/CTAB on lysozyme [144], metalloporphyrins on insulin [86], SSMs-ectoine and mannosylglyceramide (MGA) on [119], and so forth. Though all additives affect the assembly process of amyloid, different additives act through different mechanisms on amyloids. Some additives affected the whole assembly process, whereas others acted at specific assembly stages.

Cho et al. [24] reported that the anionic lipid PS stimulated amyloid aggregation only at a certain stage. ScFv-6E enhanced the kinetic aggregation of httex1-51Q by binding and stabilizing the nascent fibrils which reduced the thermodynamic lag-time of fibrillogenesis [136]. Marcus et al. [145] suggested that the isolated scFv possibly targeted a shared fibrillar motif which might be the cross-β-sheet characteristic of amyloid fibrils. Further investigation suggested that those bonds appeared after lag-time stages. The random coil to β-sheet conformational transition of Aβ was rapidly accelerated by Trimethylamine N oxide and glycerol [82], but the final stage of amyloid formation was dominated by osmolyte-facilitated changes in Aβ hydration. Some additives function as inhibitors in amyloid aggregation. Cholesterol sequestered the amylin aggregation [24], metalloporphyrins inhibited insulin aggregation [86], and chelator of Zn induced a slow but nonfibrillar aggregation of globular Aβ [142]. Kanapathipillai et al. [119] suggested a preferential exclusion mechanism of amyloid aggregation by adding denaturing agents. In their study, mixtures of ectoine and MGA, hydroxyectoine, and MG were employed to affect amyloid formation process. The results indicated that the former could inhibit amyloid formation whereas the latter could not. They found that hydroxyectoine and MG, respectively, possessed more hydrophilic features and negative charges because of their carboxyl group. In addition, , consisting of N-terminal polar heads and long hydrophobic tails, seemed to only interact with its polar head in most hydrophilic solutes. It was found that Aβ N-terminal hydrophilic domains could disassemble amyloid fibrils [116]. Similarly, mature fibrils could be disassembled by a cationic gemini surfactant, C12C6C12Br2 micelles, in vitro [115]. Synergistic, hydrophobic, and electrostatic interactions are responsible for the disassembling of fibrils. Tang et al. [146] reported the assembly-disassembly processes of α-synuclein (α-Syn) fibrils in different solutions and chaotropic agent guanidinium chloride rapidly breaking the long α-Syn fibrils into fragments.

3.4.7. Substrate Effect

AFM-based experiments indicated that identical species are apt to form different morphologies [52] at different rates [147] on distinctive substrates such as mica, graphite, gold, glass, lipid membranes, and cell surfaces. Growth rate, orientation, and deformation of the aggregation were greatly influenced by the substrates used in the experiments. Some correlations exist between substrates and amyloid fibrils conformation [4, 11, 16, 29, 30, 42, 99, 107, 108, 135, 148, 149]. More and more results attributed this phenomenon to intermolecular interaction of static electronic interaction between amyloids and substrates [130, 135, 137, 150, 151]. Linear structures and uniform elongated sheets formed on mica and graphite substrates, respectively. At the same time, the orientation of assembled sheet structures can also be affected by substrates. The amount of fibrils, found in AFM images [148], suggested that the aggregation rate of protein covalently immobilized on a silicon surface was 4.6 times faster than that on a gold surface. Distinct hydrophilic and hydrophobic conformations formed on corresponding solid substrates. Right-handed helical orientation of beaded fibrils [47] formed on a hydrophobic interface, while left-handed helical orientation formed on a hydrophilic mica surface. However, Aβ42 could not form fibrils on the surface of planar lipid. The reason might be that mica is crystalline and has negatively charged surfaces but the lipid membrane has a soft and fluid nature. Zhang et al. [42] reported that , respectively, formed large-scale, highly ordered, parallel-oriented surface patterns on different lipid membranes. Their observation implied that the properties of lipid membranes, such as the fluidity, were associated with the parallel-oriented fibrogenesis. Wegmann et al. [30] and Karsai et al. [99] also reported that heterogeneous shapes occurred on cell surfaces. Kiselev et al. [148] revealed the deformation of protein molecules immobilized on mica surfaces, and they reported that some species were preferentially adsorbed on specific substrate defects, such as edges of defects. Through myoglobin (Mb) adsorption on ZrO2-P substrate, Bellezza et al. [131] found that ZrO2-P nanoparticles affected the morphology and the interaction which resulted in prefibrillar-like aggregates. Furthermore, Liang et al. [61] found that different-staged Aβ had distinctive mechanisms of aggregation. At present, it is hypothesized that hydrophobicity is the main driving force of Aβ and liposome interaction.

3.4.8. Other Disturbances’ Effect

Other factors, such as stirring, dehydration, and magnetic fields, were found to have effects on the amyloid aggregation as well. Stirring accelerated the formation of amyloid fibrils [147]. The internal structure of fibrils was changed by dehydration [152]. Hill [153] reported that aligned aromatic peptide tubes formed in strong magnetic fields, which benefited the fibril growth [139].

4. Intramolecular Interaction

Apart from the studies using AFM imaging to investigate intermolecular interactions, the intramolecular interactions were explored by AFM-based single-molecule force spectroscopy (SMFS). SMFS has been employed to probe the mechanical properties of various biological molecules, such as polysaccharides, DNA, and proteins. A unique mechanical response representing the fingerprint of the corresponding molecule was discovered. For example, the length transitions in the mechanical fingerprint of polysaccharides were attributed to the shift of individual pyranose rings from chair to boat or inverted chair conformations [154, 155]. The extensive conformational change of a B-S transition was observed in stretching dsDNA. Moreover, direct measurement of intramolecular interactions, including donor-acceptor, ionic, conjugational, and hydrophobic interactions, has been performed. Recently, various amyloid fibrils were probed. The mechanical properties of β-sheets were gradually observed with SMFS.

4.1. Amyloid Fibrils’ Unfolding

Force measurement of amyloids focused on the β-sheet structures existing in and peptides [29], α-synuclein [64], [156], unicellular Subaerial Algae [122], terrestrial alga Prasiola linearis [83, 157], and glucagon [117]. Figure 4 illustrates the unfolding mechanical signatures of α-Syn (there are three tandem titin I27 domains on either side of the α-Syn sequence [64] in Figure 4(a)). The repetitive saw-tooth patterns during the stretch process represent the typical mechanical response of multidomain proteins of titin [158]. During the stretch process, the increasing and the abrupt force drop in each saw-tooth pattern reveal that one I27 domain was stretched and unzipped. So, the six peaks on left side with identical spacing and amplitude indicate the regular inner structure of β-sheet (Figure 4). The last peak in the saw-tooth pattern corresponds to the detachment activity between the molecule and the tip. The number of unzipping peaks agreed with that of the I27 domains composing the protein, and the indistinguishable peaks suggested a series of identical structures. The spacing gap between each saw-tooth pattern in figure is 28 nm for tandem titin I27 domains, and the approximate force value of the six unzipping peaks is 200 nN.

For other amyloid fibrils, the force patterns exhibit different spacing and rupture forces. The saw-tooth peaks were regularly spaced with a separation of approximately 36 nm for unicellular Subaerial Algae,  nm for cement of the barnacle Amphibalanus amphitrite [159],  nm for terrestrial alga Prasiola linearis [157],  nm for Prasiola linearis [83], and  nm for glucagon [117]. The average magnitude of the force peaks of terrestrial alga Prasiola linearis was found to be  pN at the stretching rate of 2.5 to 3.0 μm/s,  pN for glucagon at the loading rate of 2 μm/s, 3.5 nN for stretching cement, and 20 pN for TTR fibrils at the loading rate of 30 nm/s. The magnitude of the force, at this extension rate, would be the characteristic of the previous systems containing hydrogen-bonded β-sheets. Each jump of the saw-tooth response was attributed to a “sacrificial bond” and “hidden length” [160]. The fingerprint of the force responses could be used to analyze the specific structure present in heterogeneous conformations [87]. In the study of stretching and peptides, staircase-like force patterns were obtained. Kellermayer et al. [29] found that the force curves for two kinds of amyloid fibril were qualitatively similar. Comparing the statics data of mechanical response, and exhibited the characteristics of the smallest plateau forces of  pN and  pN, respectively. They suggested that the smallest force was the very unit for superimposing force pattern.

4.2. Mechanical Measurement

During the pulling process of SMFS, the mechanical response can be indirectly used to measure the semiflexible properties of molecules. Their quantification was performed by fitting a Worm-Like Chain (WLC) model for the semiflexible properties. The studies revealed the mean persistence length of  nm for Coccomyxa sp.,  nm for Glaphyrella trebouxiodes [122],  nm for barnacle cement [159],  nm for [29],  nm for α-Syn [64], 0.57 for adhesive nanofibers [157],  nm for EPS [83], and  nm for glucagon [117]. Two kinds of Subaerial Algae with strong attachment to anthropogenic surfaces were selected to investigate the nanoscale adhesive properties by SMFS [157] technology. The mechanical data shows how amyloid provides cohesive strength to the adhesives, and this intrinsic mechanical property can be used to explain the attachment of these subaerial microalgae onto various surfaces in urban environments.

The stiffness of nanoscale structures was quantified using force indentation curves [86, 159, 161, 162]. By fitting indentation data, typically the mechanical Hertz model [74], Young’s modulus of the material could be obtained.

Beside the useful modulus property of amyloid, the basic force-distance curve can also provide rich information of samples. The reproducibility of the saw-tooth pattern when successive curves are taken at the same locations [64, 83, 122, 157] is strong point of view to prove that amyloid fibers are able to reassemble after being stretched. Dong et al. [117] suggested that the observed elasticity was due to a force-induced conformational transition and the reversibility was attributed to the β-helical conformation of protofibrils which allows a high degree of extension.

Insulin fibrils exhibited a nearly elastic response to the compressive load which suggested lower packing density in amyloid fibrils [161] than that in protein crystals. The measured lower Young modulus indicated that insulin fibrils possess a looser internal packing compared to globular protein crystals and agree with the loose structure of β2-microglobulin amyloid [163].

4.3. Effects of Experimental Conditions

Force responses are, in a similar fashion as the morphology, heavily influenced by experimental parameters such as loading rate, ionic concentration, pH value, and incubating time. Time-lapse AFM imaging and force spectroscopy have been performed to study the assembly process of fibrils under different experimental conditions in situ. α-Synuclein, amyloid β-peptide (Aβ), and lysozyme were used to explore the pH value influence on interprotein interaction of amyloid aggregation [87]. It has been confirmed that the pH value for these conformational transitions coincided with pH values that led to changes in the pulling forces. The SMFS data showed that the attractive force between homologous protein molecules was minimal at a physiological pH value and increased dramatically at an acidic pH value. However, it has not been directly proven that the dramatic increase in interprotein interaction under acidic conditions was responsible for fibrillation.

5. Summary and Outlook

We reviewed the latest observations of inter- or intramolecular interactions of amyloid fibrils using AFM and AFM-based SMFS techniques. Various morphologies of amyloid fibrils, the assembly process, and the aggregating pathways were summarized in order to analyze their influence on amyloid fibrillogenesis. In addition, the fingerprint of mechanical response through AFM-based SMFS complements the information gained by topological AFM imaging. There is no doubt that SMFS combined with AFM provides a useful application in detecting inter- or intramolecular interactions. They opened a new path to explore fibrillogenesis, provide information of amyloid fibrils, and finally initiate a solution to curing neurological disordered diseases.

Competing Interests

The authors declare that they have no competing interests.


The authors acknowledge financial support from the National Natural Science Foundation of China (nos. U1304310 and 21373077), and Henan Natural Science Research Office of Education project (no. 12A180005 and no. 2010A140001).