Illuminating amyloid fibrils: Fluorescence-based single-molecule approaches.

The aggregation of proteins into insoluble filamentous amyloid fibrils is a pathological hallmark of neurodegenerative diseases that include Parkinson's disease and Alzheimer's disease. Since the identification of amyloid fibrils and their association with disease, there has been much work to describe the process by which fibrils form and interact with other proteins. However, due to the dynamic nature of fibril formation and the transient and heterogeneous nature of the intermediates produced, it can be challenging to examine these processes using techniques that rely on traditional ensemble-based measurements. Single-molecule approaches overcome these limitations as rare and short-lived species within a population can be individually studied. Fluorescence-based single-molecule methods have proven to be particularly useful for the study of amyloid fibril formation. In this review, we discuss the use of different experimental single-molecule fluorescence microscopy approaches to study amyloid fibrils and their interaction with other proteins, in particular molecular chaperones. We highlight the mechanistic insights these single-molecule techniques have already provided in our understanding of how fibrils form, and comment on their potential future use in studying amyloid fibrils and their intermediates.

Introduction

Most proteins must fold into a specific 3D conformation in order to achieve (and maintain) their functional and active state [1]. During the folding process the polypeptide chain typically samples numerous conformations until the most energetically favourable fold is reached [1], [2]. Under conditions of cellular stress (e.g. changes in pH or temperature) or as a result of genetic mutations, proteins may not fold correctly or become misfolded, resulting in hydrophobic regions normally buried within the interior of the protein becoming exposed on the surface [3], [4]. Misfolded proteins can be refolded or discarded by cellular protein-quality control mechanisms that include molecular chaperones [5], [6] and degradation pathways (such as the ubiquitin proteasome pathway and autophagy).

If protein-quality control mechanisms are (or become) ineffective, misfolded proteins can self-associate to form amorphous or fibrillar aggregates [3]. Fibrillar protein aggregates, in the form of highly ordered, β-sheet rich ‘amyloid’ fibrils, have received much attention due to their association with various diseases, including neurodegenerative disorders such as Alzheimer’s Disease (AD), Parkinson’s disease (PD), and Amyotrophic Lateral Sclerosis [3], [7], [8], [9]. These fibrillar protein aggregates and the mechanism by which they form are the focus of this review. Although the presence of amyloid fibrils within insoluble, proteinaceous inclusions has been widely documented [10], [11], mechanistic details regarding how fibrils form and cause harm to cells remain to be elucidated. This is due, at least in part, to the complex and heterogeneous nature of the fibril-forming process and the oligomeric state of the aggregated protein.

Many of the assays traditionally employed to measure amyloid fibril formation utilise bulk biochemical techniques, which are based on the ensemble-averaging of a large number of molecules in solution. These assays typically report on β-sheet formation (indicative of fibril formation) and any ensemble-averaged changes to this process due to manipulation of the surrounding environment. Single-molecule techniques are valuable methods to study the heterogenous biological process of fibril formation as they can unmask mechanistic details “hidden” by ensemble-based measurements [12], [13]. Single-molecule techniques comprise an array of powerful methods that include cryo-electron microscopy, optical tweezers, mass photometry [14] and fluorescence-microscopy approaches [13], [15]. Single-molecule fluorescence-microscopy techniques can provide mechanistic detail about molecular interactions such as enzymatic turnover rates, dissociation and association rates, and stoichiometries of interacting species [16], [17], [18]. As such, they are considered ‘bottom-up’ methods in that by visualising and analysing information from single molecules, they allow information about an entire system to be ascertained. Such approaches can be particularly powerful when applied to biological processes that depend on rare or transient intermediate species/states in order to proceed. One such process is the formation of amyloid fibrils, wherein these species are often masked by ensemble-averaging methods. This review focusses on the use of fluorescence-based single-molecule techniques to study amyloid fibrils: their formation, disaggregation, and the interactions that govern these processes.

We first compare how Total Internal Reflection Fluorescence (TIRF) and confocal microscopy can be utilised as single-molecule fluorescence methods, before outlining the techniques of stepwise photobleaching analysis, Förster Resonance Energy Transfer (FRET), two-colour coincident detection (TCCD) and fluorescence correlation spectroscopy (FCS). We review how each of these methods have been used to study amyloid fibril formation at the single-molecule level, and highlight their future potential in understanding the molecular mechanisms that drive this process. We conclude by discussing single-molecule techniques that have been employed to investigate how molecular chaperones interact with amyloid fibrils to inhibit their formation and facilitate their disaggregation.

Oligomerisation and kinetics of fibril formation

The formation of amyloid fibrils occurs via a nucleation-dependent pathway in which partially folded intermediate states of proteins self-associate as a result of surface-exposed hydrophobic regions. Initiation of amyloid fibril formation and structural polymorphisms within fibrils can be triggered by factors such as genetic mutations, intracellular protein concentration, specific proteolytic cleavage and ligand binding [19]. The self-association of misfolded amyloid-precursor proteins promotes β-sheet stacking which stabilises core amyloid fibril structures [20]. This stability is facilitated by backbone hydrogen bonding between β-sheet strands of self-associated protein monomers [20], hydrogen bonding between the loops connecting β-sheets, interactions between polar and charged side chains, and interlocked β-sheets forming ‘dry’ steric zippers [21], [22], [23]. The ability of dyes such as Thioflavin T (ThT) to bind β-sheet structures [24], [25] facilitates the study of amyloid fibrils by bulk biochemical assays and some single-molecule fluorescence techniques. These dyes typically become fluorescent once bound to β-sheet structures present in amyloid fibrils [128] and using these dyes to monitor the fibril-formation process reveals the presence of a lag phase, elongation phase, and plateau (indicating equilibrium) in the reaction [26] (Fig. 1). The lag phase is the rate-limiting step of the reaction and represents the time taken for β-sheet rich oligomers to form a sufficient percentage of nuclei or ‘seeds’ [27]. During the lag phase, monomeric and early-stage oligomeric species are dominant in solution, resulting in low, mostly undetectable levels of fluorescence emitted by amyloid dyes [28], [29]. The association of monomers with these nuclei results in their extension and the characteristic increase in fluorescence observed during the elongation phase. The plateau phase represents the stage of the reaction when an equilibrium has been reached between all species in solution [30] (Fig. 1).

Fig. 1

The formation of amyloid fibrils as monitored by ThT fluorescence. This schematic describes the formation of amyloid fibrils from monomeric protein subunits. The lag phase represents the association between β-sheet rich monomers to form oligomers and is the rate-limiting step in the process. Production of oligomeric seeds is followed by the elongation phase during which fluorescence increases due to the rapid addition of monomers to oligomers. When an equilibrium is reached the fluorescence plateaus. Mature amyloid fibrils can facilitate secondary nucleation on their surface or undergo fragmentation. Both events generate oligomers that can further propagate fibril formation.

Amyloid fibrils can consist of up to six pre-fibrillar filaments, called protofibrils, which intertwine to form more rigid, straight (mature) fibrils 5–20 nm in width [31], [32], [33]. Whilst fibrils typically share a common underlying β-sheet rich architecture (initiated by oligomer formation during nucleation), there have been multiple polymorphisms of mature amyloid fibrils formed by the same protein detected by cryo-electron microscopy (EM), x-ray crystallography, and nuclear magnetic resonance (NMR), including ribbons, sheets and multi-strand cables [34]. It has been suggested that the heterogeneity of fibril and oligomer structures is related to peptide-peptide interactions that mediate formation of β-sheet twists [35]. Recently, cryo-EM technology has resulted in an increase in structural detail surrounding amyloid fibril polymorphisms. The complexity and heterogeneity of amyloid fibrils has been reviewed elsewhere [21], [36] and so are only briefly discussed here.

AD [37] and PD [38] are associated with amyloid fibrils formed by the proteins Amyloid β (Aβ) and α-synuclein (α-syn), respectively. It has been found that in both cases, fibrils form by the interaction of two protofibrils via a common dimer-interface, a feature that has prompted further research into their structural characteristics [37], [38]. Structural elucidation of α-syn fibrils has shown a characteristic left-handed twist morphology, extensive Greek-key topology, and that most of the residues corresponding to the genetic mutations responsible for early onset PD (i.e. the point mutations A53T [9], E46K [39] H50Q, and G51D [40]) are located in the protofilament dimer interface or involved in protofilament stabilisation [38]. Brain-derived Aβ amyloid fibrils have a right-handed supertwist morphology [41], and thus have a major structural difference from the left-handed twist formed by in-vitro formed α-syn and Aβ fibrils. The array of fibril morphologies highlights the inherent heterogeneity of amyloid fibrils and the value in developing and utilising single-molecule techniques to study them.

The oligomeric intermediates formed during the process of fibril formation have received considerable attention as these are considered to be the most cytotoxic species [42], [43], [44], [45]. These oligomers are heterogeneous, both in terms of their structure and propensity to mature into fibrils [45], [46]. The formation of oligomers can occur via primary nucleation as a result of the assembly of monomers through β-sheet interactions [47] and can then act as ‘seeds’ for the formation of protofibrils and fibrils. Toxic oligomers can also form through processes of ‘secondary nucleation’ and fragmentation of mature fibrils. Secondary nucleation is catalysed by the surface of pre-existing fibrils [48], [49], [50], [51], [52]. The properties of the ‘parent’ fibril that facilitate secondary nucleation remain to be established; however, it has been postulated that hydrophobic patches on the fibril surface [53], and the ability of misfolded monomers to ‘recognize’ similar structures within the ‘parent’ fibril play a role in this process [54]. It was recently shown that decreasing the surface hydrophobicity of fibrils formed by the AD-associated Aβ(1-42) via serine substitution of hydrophobic residues exposed on the fibril surface, does not change the rate of secondary nucleation [53]. However, both α-syn [55] and Aβ(1-42) [53] are able to ‘template’ secondary nuclei via the exposed surface residues of the mature fibrils, wherein monomers transiently bind and form oligomers with identical morphologies to those of the templating fibril. These results suggest that structural features of fibrils, as well as the solution conditions under which they are formed [55], facilitate secondary nucleation and thus increased toxicity [54], [56].

The fragmentation of mature fibrils also results in the formation of cytotoxic oligomers [51], [57], [58], [59], [60]; this is a length-dependent process whereby the stability (per monomeric unit) of fibrils decreases with increasing length and thus promotes their fragmentation into shorter oligomeric units [61]. Thus, fragmentation increases the concentration of prion-like ‘seeds’ in solution, in turn promoting rapid fibril formation and the spread of toxic oligomers between tissues [59], [60], [62]. Furthermore, fragmentation has been implicated in the prion-like behaviour of amyloid oligomers in disease, as demonstrated by both experimental [60], [63], and theoretical [59], [60], [62] investigations. Both secondary nucleation and fragmentation are widely acknowledged as significant contributors to the exponential phase of amyloid fibril formation [52], [57], [59], [64], [65]; however, due to the difficulty in characterising such dynamic processes using ensemble-based techniques, these processes have not yet been directly observed or fully characterised.

Amyloid fibril formation can be reversed through the process of disaggregation, wherein fibrils return to a soluble state via the release of monomers and oligomers back into solution. The disaggregation of amyloid fibrils often involves molecular chaperones, their co-chaperones, and essential nucleotide exchange factors (NEF). The most widely studied fibril disaggregation system is the Hsp70 ‘disaggregase’, which involves a combination of Hsp70, Hsp40, and Hsp110 (a NEF) proteins [66], [67]. This chaperone system is able to interact with mature amyloid fibrils, driving their disassembly back into a soluble state [68], [69], [70]. As such, the Hsp70 disaggregase has attracted interest as an avenue to combat neurodegenerative disease by potentially reversing the aggregation process [66], [71], [72], [73], [74], [75].

Single-molecule techniques

Despite the advances that have been made toward understanding amyloid fibrils there are still gaps in our knowledge, especially when it comes to the mechanisms responsible for the complex processes of secondary nucleation, fragmentation, and disaggregation. Largely this gap in knowledge is attributable to the complexity of studying such heterogeneous species with ensemble-based biochemical techniques. Single-molecule methods have great potential for studying individual amyloid fibrils, and fluorescence-based single-molecule approaches have the capacity to reveal the mechanistic details that underpin the dynamic processes of secondary nucleation and fragmentation [76], [77], [78], [79]. For example, the single-molecule technique of electron microscopy [41], [80] has been instrumental in describing the structural features of amyloid fibrils including the many polymorphisms between protofibrils, mature fibrils, and oligomers [81], [82]. Structural variations that have implications for in vitro research in this field have been determined by electron microscopy, with a notable example being the aforementioned differences between brain-derived and in vitro amyloid fibrils [41], [83]. However, structural-based single-molecule techniques are limited in their ability to capture the dynamic nature of the fibril-forming process.

Fluorescence-based single-molecule techniques, wherein fluorescent moieties are attached or associate with proteins so that they can be observed, enable dynamic processes to be captured and analysed. For example, single-molecule fluorescence methods have been used successfully to describe the three-state folding process of a single protein molecule [84], fluctuating enzymatic reaction rates of single cholesterol oxidases [85], and the multiple intermediate states present in the lag phase of amyloid formation [86]. Fluorescence-based single-molecule techniques involve using optical methods to monitor fluorescence probes conjugated to biological molecules, such as proteins [87]. This includes the use of dyes that become fluorescent upon association with the structures of amyloid fibrils [88], [89], and small fluorescent tags (fluorophores) that can be covalently linked to proteins. Fluorescence from these probes can be monitored at low (pico- or nano-molar) concentrations in solution using confocal microscopy, or immobilised on a surface in TIRF-based imaging (Fig. 2). It is well established that protein aggregation can be dependent on protein concentration, both in vitro [90], [91] and in vivo [91]. Monitoring protein aggregation often requires the use of relatively high (micro-molar) concentrations of protein in order to study the mechanisms of aggregation on experimentally accessible timescales. However, the relatively high concentrations of protein required potentially limits the real-time study of protein aggregation by single-molecule fluorescence methods since these approaches typically require very low concentrations of the molecule of interest (in order to reduce background fluorescence originating from the excited volume). Thus, kinetic studies of amyloid fibril growth by single-molecule fluorescence microscopy typically require samples to be aggregated at higher concentrations and then diluted (to pico- or nano-molar concentrations – chemical crosslinking agents can be used to stabilise interactions during this process) for visualisation.

Fig. 2

Setup and comparison of fluorescence-based single-molecule microscopy techniques. (A) The schematic shows the principal components required for confocal microscopy. An excitation laser is directed through a series of mirrors, confocally aligned pinholes, and an objective lens onto a small area of the sample up to a depth of ~ 500 nm. Photons emitted from the excited samples that pass through the focal volume are detected by photon detector arrays for analysis. (B) For TIRF, the laser light is directed at the surface of a coverslip, where total internal reflection occurs by an evanescent wave, allowing emission of light from fluorophores within ~ 100 nm from the coverslip-solution interface.

Solution-based versus surface-immobilised fluorescence single-molecule approaches

Depending on the questions asked and experimental design constraints, two different fluorescence-based approaches are typically employed to study the dynamics of biomolecules at the single-molecule level: confocal methods that detect fluorescent species freely diffusing through solution and transiently passing through a tightly-focused excitation volume, or TIRF-based techniques that monitor through time a fluorescent species that is held stationary by immobilisation to a surface (Fig. 2).

Confocal microscopy enables the visualisation of fluorescently labelled biomolecules by the creation of a diffraction-limited excitation volume inside a sample using light of a wavelength that coincides with the excitation maximum of the fluorophore to be monitored [92], [93] (Fig. 2A). This excitation light is passed through confocally aligned pinholes, which results in a small volume (200 nm × 200 nm × 500 nm in dimension) being illuminated, and the emission light from that volume being detected while excluding out-of-focus sources of signal [93], [94]. The traditional approach to confocal microscopy utilises ‘optical sectioning’, whereby the laser is scanned through a sample in all three spatial dimensions (depths of up to 500 nm can typically be achieved), and a three-dimensional image of the sample (such as a cell) is obtained [95], [96].

An alternative use of a confocal microscope is to place the confocal volume inside an aqueous solution, detecting the transient bursts of fluorescence originating from individual labelled species as they diffuse through the excitation volume [97]. Following excitation of a fluorophore, photons emitted from the molecule can be detected (typically using single-photon detector arrays), thus providing single-molecule sensitivity [96], [98]. In contrast to TIRF microscopy (Fig. 2B) (discussed below), the ability to place the confocal volume deep inside a three-dimensional sample allows for molecules well above the solid–liquid interface to be visualised [99]. Subsequently, this type of confocal microscopy avoids having to immobilise molecules to a coverslip surface and thus avoids problems with steric hindrance that can arise with the TIRF microscopy approach. Single-molecule confocal microscopy enables information to be obtained on freely diffusing, fluorescently labelled molecules. Common methods that employ this approach include (i) fluorescence correlation spectroscopy (FCS), in which the apparent sizes of molecules as they pass through the confocal volume are calculated by analysis of their diffusion, and (ii) two-colour coincidence detection (TCCD), in which the simultaneous detection of two fluorescent signals within the confocal volume is used to detect biomolecular associations. Experiments that couple confocal microscopy with time-resolved single-photon detection enable the collection of data on fluorescence lifetimes [96], and fast dynamics via Förster resonance energy transfer (FRET) based applications [96], [100], [101] vastly increase the information that can be obtained from single-molecule confocal imaging approaches. A limitation of confocal microscopy is the size of the confocal volume: molecules can only be monitored for the period of time they remain in the confocal volume (often data acquisition has to occur on the nano- to milli-second time scales) [102].

Monitoring of single-molecule dynamics over extended time scales can be achieved using a TIRF microscopy approach (Fig. 2B). This technique involves directing the laser used for excitation of the fluorophore onto a liquid–solid interface (such as a quartz coverslip) at an angle such that the light is internally reflected. This produces an evanescent wave within the solvent at the liquid–solid interface (between the coverslip and the sample-containing aqueous solvent on its surface) [99], [103]. The evanescent electro-magnetic field (created by laser excitation) decays exponentially in intensity and thus is confined to close proximity to the surface (~100 nm) (Fig. 2B). This allows selective excitation of molecules on the surface while effectively suppressing background fluorescence from other molecules in solution [103], [104]. In this way TIRF can provide accurate, detailed information regarding molecular interactions [105]. Fluorescence-based TIRF methods, such as FRET (including single-molecule FRET) and photobleaching (to determine stoichiometries of interactions via the stepwise photobleaching of fluorophores) can be exploited to determine the molecular details of protein–protein interactions. Moreover, the use of fluorescent probes, such as aggregation-induced emission (AIE) dyes, in combination with TIRF-based single-molecule approaches has facilitated the visualisation and tracking of protein aggregates without the need for a fluorophore conjugated to the protein of interest [106], [107], [108].

Fluorescent probes

Aggregation-induced emission (AIE) dyes

Aggregation-induced emission has been demonstrated to be advantageous for the visualisation of amyloid fibrils by TIRF microscopy (Fig. 3A) [24], [88], [109]. Single-molecule techniques based on non-covalent interactions between the probe and protein target avoid issues that can occur when fluorescent proteins are chemically modified for fluorescent labelling (Fig. 3B). This is particularly relevant to studies of the aggregation of α-syn as it has been reported that the fibril length and morphology can vary when the protein is tagged with fluorescent proteins or fluorophores [110]. Moreover, covalent labelling of proteins with a fluorophore can be problematic due to photobleaching that occurs during imaging. This bleaching restricts the time available for imaging and limits the timeframe for kinetic studies due to the decrease in signal intensity [111].

Fig. 3

Fluorescence-based approaches for monitoring amyloid fibril formation of proteins. (A) AIE dyes bind to stacked β-sheet structures within proteins, facilitating the detection of amyloid fibril formation. The signal from an AIE dye increases with β-sheet structure and is therefore used as an indicator of amyloid fibril formation. Fluorescence from AIE dyes does not photo-bleach as free dye molecules exchange with those bound to the fibril, maintaining the signal over time. (B) Fluorophores that emit light at various wavelengths can be directly conjugated to individual protein subunits, such as monomers, that make up amyloidogenic oligomers and fibrils. Photobleaching of each covalently attached fluorophore can be used to determine the number of subunits in a protein complex or fibril. In each case, emission from the fluorophore through time can be monitored by confocal or TIRF microscopy.

Fluorescent dyes including Nile Red [112] and ThT [113], [114], have been used to image amyloid fibrils in TIRF studies as their non-covalent, transient interaction avoids issues with photobleaching yet still allow direct observation of protein aggregates (Fig. 3). This sustained fluorescence is facilitated by the constant exchange of dye molecules between those in solution and those bound to the amyloid structures, minimising any loss of signal over time. For example, Nile Red allows observation of exposed hydrophobic regions [115], [116], as the presence of highly polar regions on proteins or lipids stimulates structural changes in the dye [117]. Specifically, changes in polarity result in altered stability of the dipole moment created by twisted intramolecular charge transfer between the donor and acceptor regions of the dye [118]. Consequently, a shift in emission wavelength [115] is observed. This phenomenon has been exploited by tracking the shift in emission wavelength of Nile Red bound to the surface of α-syn amyloid fibrils, allowing surface hydrophobicity of the fibrils to be mapped and thus the surface characteristics of the pathological protein aggregates to be determined [112], [119].

The AIE properties of ThT have resulted in its application for the visualisation of amyloid fibrils [24], [25], [112], [113], [114], [120]. An important parameter in describing the ability of a molecule to emit fluorescence is the quantum yield, defined as the ratio of photons emitted to those absorbed and reported as a number between zero and one. When free in solution, ThT molecules display rotation around a single C—C bond, resulting in a non-fluorescent ‘dark’ conformation, and a quantum yield approaching zero [121]. The binding of ThT to stacked β-sheets in amyloid fibrils restricts this rotation such that the absorbed light is emitted, thus increasing the quantum yield to ~ 0.44 [24], [25], [121], [122], [123], [124]. The fluorescence of ThT is relatively sensitive to the presence of amyloid fibrils due to its propensity to intercalate within β-sheet structures that are not present until oligomeric seeds and/or fibrils have formed; thus, for most proteins a fluorescent signal is not detected when the primary species in solution are monomers [24]. However, ThT has limitations with regards to using it as a fluorescent probe for single-molecule studies of protein aggregation. The relatively small Stokes shift of ThT (the emission photon wavelength is close to the wavelength of the excitation photon [125]), can cause problems in separating the emission signal from scattered excitation light [126], [127]. Recently, AIE dyes with larger Stokes shifts that can be used to visualise fibrils have been described [29], [108]. Apart from the advantage of an increased affinity for amyloid fibrils, another advantage of some of these recently described AIE dyes is that two-colour experiments can be performed using only one excitation laser. In these experiments, amyloid fibrils labelled with one such AIE dye, and a protein labelled with a fluorophore of a spectrally separate emission wavelength (but the same excitation wavelength) would be present. This experimental scenario would thus only require one excitation laser, whilst still being able to identify both labelled species in the sample via their different emission wavelengths. As an example, the AIE dye α-cyanostilbene derivative ASCP [29] has a very large Stokes shift (145 nm) and can be excited using a 488 nm laser light source which is commonly found in commercial fluorescence microscopes (excitation maximum at 460 nm and emission maximum at 605 nm). Recently, ASCP was used to simultaneously image molecular chaperones (labelled with an AlexaFluor488 fluorophore and emission monitored at 500–550 nm) as they interacted with α-syn fibrils (excited with the 488 nm excitation wavelength and emission monitored at 675 nm and beyond) [108].

Conjugation of fluorophores to proteins

Fluorophores that covalently label proteins (Fig. 3B) are ideally small, bright, stable, and do not interfere with the protein structure or function. Generally, labelling of protein monomers with a small-molecule fluorophore gives the researcher flexibility with regards to the location and type of tag (optimised methods on this approach are reviewed in depth elsewhere, e.g. [128]). However, it is important to note that it is not always possible to attach a fluorophore to a protein without modifying its structure or function. As an example, covalent attachment of fluorophores to Aβ(1-42) has been shown to alter both the molecular weight of oligomers and the structure of mature fibrils formed by the protein, depending on the hydrophobicity and pH dependence of the fluorophore used [129].

Whilst the small oligomeric species formed early during amyloid assembly are considered to be the most pathogenic [130], they have been technically challenging to investigate using AIE dyes and bulk-ensemble approaches. This is because these oligomers are often transient and without extended β-sheet content and thus the overall fluorescent signal from these dyes in solution is very low. Studies looking to characterise these small oligomers using single-molecule fluorescence methods have typically exploited fluorophores covalently attached to the protein of interest. Using such an approach, the size of oligomeric Aβ(1-42) species was determined using TIRF microscopy and stepwise photobleaching of the attached fluorophore [131](described in more detail in Section 3.3.1). This enabled the size of Aβ(1-42) oligomers to be ascertained, and comparisons drawn between those formed in the absence and presence of putative inhibitors of fibril formation. A similar approach has been used to monitor the disaggregation of amyloid fibrils into smaller subunits (including monomers) [132], [133], [134]. Single-photon detection on a confocal microscope has been exploited to monitor the inhibition of oligomerisation and the disaggregation of pre-formed α-syn fibrils in which individual α-syn monomers were labelled with AlexaFluor-488 fluorophores [133]. By monitoring the changes in fluorescence of α-syn, this approach enabled the screening of a range of polyphenol compounds for their capacity to disaggregate the α-syn fibrils [133], [134]. The mechanistic insights these single-molecule experiments provide cannot be obtained through ensemble-averaging approaches alone, thus highlighting the crucial importance of single-molecule methods in identifying regulators of fibril formation.

Fluorescence-based single-molecule techniques to study amyloid fibril formation

Single-molecule fluorescence approaches such as photobleaching, TCCD, FCS, and FRET are commonly used to investigate protein aggregation by monitoring the emission of fluorophores bound to the protein of interest (Fig. 4). These powerful techniques are capable of probing dynamics and interactions of the proteins to which they are bound.

Fig. 4

The single-molecule methods of two-colour coincidence detection (TCCD), fluorescence correlation spectroscopy (FCS) and Förster resonance energy transfer (FRET). Proteins that are conjugated to a fluorophore emit fluorescence at the emission wavelength specific to the fluorophore label. (A) TCCD: Molecules diffusing through the confocal volume that change in size over time (e.g. during amyloid fibril formation or disaggregation) result in fluorescence burst intensities proportional to their size, and a change in the diffusion rate. These are analysed and plotted as frequency distributions for each population present. (B) FCS: The fluctuations in fluorescence observed over time can be analysed using statistical correlation functions to precisely determine changes in size of a molecule over time. (C) FRET: Proteins labelled with two FRET competent fluorophores may change in conformation, bringing the fluorophores closer together which facilitates a donor fluorophore (orange) transferring energy to an acceptor fluorophore (blue), resulting in detectable emission from the acceptor and thus an increase in FRET efficiency. In this way, the FRET efficiency can report on conformational changes in the protein. These FRET signals can be observed whilst the protein is immobilised on the surface of a coverslip for TIRF microscopy or diffusing through a confocal volume for confocal microscopy.

Photobleaching

Photobleaching is a photophysical phenomenon wherein fluorophores undergo permanent chemical damage caused by prolonged excitation, often due to oxidation by free radicals within the imaging solution; as a result of this damage the fluorophore no longer produces a fluorescent signal (Fig. 3B) [135], [136], [137], [138]. In the context of investigating fluorescently labelled proteins, photobleaching can be a limitation as it restricts the time available for observations and measurements. However, by exploiting photobleaching it is possible to quantitatively characterise stoichiometries and subunit copy numbers. This is achieved by counting photobleaching steps in the signal arising from a multi-protein complex as labelled proteins bleach one by one [139]. To do so, there are several experimental requirements: proteins of interest must be labelled at a high efficiency and known labelling ratio to reduce the possibility of unlabelled protein subunits being incorporated into complexes. In addition, proteins must be immobilised at low spatial density to allow for the separate monitoring of individual proteins without signal from neighbouring complexes impacting the recorded fluorescence [140]. In order to satisfy these conditions, TIRF microscopy is the most commonly used platform for photobleaching experiments [140].

Single-molecule photobleaching has been exploited to characterise heterogeneous populations of Aβ (1-40) oligomers [141]. Gel-filtration chromatographic analysis of this population did not allow for the conclusive assignment of the size of oligomers in solution [141]. By using a single-molecule photobleaching approach, the range of oligomers in solution could be characterised, including the identification of hexameric oligomers that were not detected using gel filtration [141]. Additionally, photobleaching has been used to demonstrate that Aβ(1-40) oligomers in a heterogeneous range from monomers to hexamers are able to bind to neuroblastoma cells, with the most abundantly bound oligomeric species being dimer/tetramer complexes [142], supporting the theory that smaller oligomeric species are more cytotoxic and capable of binding cells.

Whilst single-molecule photobleaching is clearly a valuable tool for characterising protein complexes, there are some aspects that require consideration, in particular in relation to protein aggregates. For example, analysis of the time-versus-intensity trajectories generated during photobleaching experiments often requires selection of trajectories that have ‘defined’ photobleaching steps, a process that can be highly subjective [143]. Improvements to analyses of this kind are being introduced with the use of deep-learning analysis pipelines to remove the subjectivity of selecting trajectories [143]. Additionally, labelled proteins are prone to ‘pre-bleaching’ wherein a proportion of fluorophores bleach prior to measurements being made. This phenomenon has been corrected for previously by creating a bleaching standard using covalently-linked fluorescently-labelled multimers [144]. The authors applied this correction to the photobleaching analysis of diabetes-associated human islet amyloid polypeptide (IAPP) in order to determine that the membrane binding affinity of amyloid peptides is in the order of trimer > dimer > tetramer ≫ monomer, in line with previous results obtained via FCS [144], [145]. A similar approach could be employed to correct for the issue of fluorophore quenching, a phenomenon wherein the close proximity and altered dipole of fluorophores in large amyloid aggregates can result in the dimming of fluorescence and thus potentially resulting in an inaccurate determination of subunit counts and stoichiometries [141].

TCCD

Two-colour-coincidence detection (TCCD) is a fluorescence-based single-molecule method that allows quantification of the number of freely diffusing molecules in solution, and the association (including stoichiometries) between molecules [146]. These two-colour measurements rely on each biomolecule being labelled with its own distinct fluorophore, such that each fluorophore can be excited by spatially overlapped diffraction-limited, confocal laser foci [101], [146] (Fig. 4A). The solution is imaged at pico-molar concentrations of fluorescent species and the signal at multiple emission wavelengths is used to detect the presence of coincident fluorescent bursts within the confocal volume; the respective fluorescent intensities can be used to determine the number of biomolecules present in a complex [77], [101], [119]. Dilution of samples down to the pico-molar range is required for TCCD as this decreases the likelihood of fluorescent species that are not part of a complex simultaneously traveling through the small probe volume [101], [147]. The requirement of very low concentrations of the fluorescent species in TCCD is an important consideration when studying transient/weak protein–protein interactions: chemical cross-linkers may be required to capture interactions for TCCD.

The disaggregation of Aβ fibrils has been tracked using TCCD, wherein the release of fluorescently labelled oligomers and monomers of Aβ(1-40) from mature amyloid fibrils was tracked over time [148]. The observed disaggregation rates from these single-molecule experiments were in accord with estimations determined previously based on hydrogen deuterium mass spectrometry data [149]. Moreover, using this single-molecule approach it has been demonstrated that both oligomers and monomers are released as Aβ fibrils disaggregate, as had been previously proposed, but had not been shown experimentally [150]. A TCCD approach was also used to determine the apparent size of α-syn oligomers and to observe the interconversion between toxic and non-toxic oligomeric forms of α- syn [146]. It was concluded that the median size of the most cytotoxic α-syn oligomeric form is 38 ± 10 molecules [146]. Cremades et al. (2012) [77] combined TCCD experiments with cell-based experiments in order to correlate the size of the oligomers with their cytotoxicity. However, it remains challenging to accurately determine fibril lengths using TCCD; this is partly due to the requirement of careful thresholding of the signal in order to define true events as opposed to background noise [101]. If this threshold is set too low, the fluctuations in noise from the instrument itself can result in false event detection [101]. Thus, the possibility arises that TCCD may not be capable of accurately determining the size of the smallest species formed early during the aggregation process [146]. In contrast, the length of smaller oligomeric species can be readily obtained via a TIRF microscopy approach, using labelled protein monomers (with one fluorophore per monomer) (as depicted in Fig. 3A). This measurement gives an absolute value of units of length within smaller oligomeric species by dividing the total fluorescence intensity by the intensity of one fluorophore, or the intensity of a photobleaching step [151]. A combination of both methods of measurement may be required to fully characterise the heterogeneous species present in solution. These quantitative measurements have been used for an analysis of the effect of oligomer size on the generation of reactive oxygen species and membrane permeability upon binding to neuronal cells [119], [123].

FCS

Fluorescence correlation spectroscopy (FCS) is a statistical technique commonly coupled to confocal microscopy which facilitates in-depth analysis of particles diffusing through the confocal volume [102], [152], [153], [154] (Fig. 4B). To perform FCS, a solution of fluorescently-labelled molecules are imaged at a concentration that corresponds to a small number [155] of molecules (~1) being present in the detection volume simultaneously: data on the transient, statistical fluctuations in the fluorescence are collected and compared to rapid changes in the fluorescence attributable to ‘noise’ in the signal [153], [156]. The fluctuations in fluorescence intensity can be analysed and assigned to a specific feature of the molecule using mathematical auto- or cross- correlation models [157]. This analysis uses the ‘diffusion time’ of a molecule to extract the size, conformation, concentration and chemical kinetic rate constants of the molecules and processes in question [157]. The hydrodynamic radius, a geometric parameter of a polymer in solution, can be used in this context as an indicator of the amount of movement of the detected proteins. An important consideration when utilising FCS is that proteins in a more spherical conformation will have a smaller hydrodynamic radii as opposed to those with a non-spherical structure [158]; whilst the hydrodynamic radius is a useful parameter for examining conformational changes, it can be difficult to ascertain whether a change in diffusion is due to a change in the molecular mass or the shape of the diffusing molecule. The complexity of the types of FCS experiments that can be performed has increased over the years, such that now proteins can be labelled with different coloured fluorescent probes in order to detect protein–protein interactions (indicative of complex or aggregate formation). Multi-colour cross-correlation analysis of bimolecular associations (referred to as FCCS, fluorescence cross-correlation spectroscopy) is useful for studying multi-component protein complexes.

Recently, Li et al. (2019) employed both FCS and FCCS to detect small differences in the size of α-syn oligomers and highlight the potential role of oligomers in the early aggregation process. Engineered tandem α-syn oligomers of specific sizes (monomers, dimers, tetramers, and octamers) were labelled in equimolar amounts with two different fluorophores, and then aggregated. At time points during the early aggregation process, FCCS was performed to determine which associations were most likely to occur [160]. It was found that early-stage oligomeric α-syn forms fibrils by self-associating with other oligomers at a faster rate and to a greater extent than with free monomers, suggesting that the well-established impact of oligomers on the rate and extent of mature fibril formation may be due to them requiring less conformational rearrangement to add to the growing fibrils compared to monomers [161], [162], [163], [164]. This result highlights the capacity of techniques with the sensitivity of FCCS to uncover mechanistic details of heterogeneous systems such as those formed during the early stages of amyloid fibril formation.

Protein aggregation can be followed by FCS by comparing diffusion rates and thus the hydrodynamic radii of species as the aggregates grow in size or change in conformation [165], [166], [167], [168], [169]. By using FCS to detect the changes in the size and conformation of early-stage Aβ(1-42) aggregates over a range of concentrations it was determined that a critical oligomerisation concentration of 90 nM exists for this peptide [170]. Oligomers with an average size of ~ 50 monomeric units and an elongated conformation, consistent with a protofibril, were observed at this critical concentration point [170]. Identifying conformational states and the concentrations at which cytotoxic oligomers form is an essential step towards developing novel therapeutics and early detection methods for diseases associated with amyloid fibril formation.

FCCS can be combined with advanced excitation regimes for improved accuracy in detecting multiple fluorescent wavelengths in the same sample. Alternating laser excitation (ALEX)-FCCS involves alternating the laser excitation wavelength on a microsecond timescale and facilitates detection of multiple fluorescent species without spectral crosstalk interfering with the accuracy of the measurements, without the need for multiple detection channels [171], [172]. This concept has been advanced by pulsed interleaved excitation (PIE) wherein the laser excitation is alternated on the nanosecond timescale, enabling multiple fluorophores to be excited within the time it takes for molecules to diffuse through the confocal volume [173], [174]. Thus, when coupled with FCCS, PIE facilitates the detection of multiple fluorophores in a single-detection channel, and removes spectral crosstalk in a multi-channel detection system which improves the accuracy of cross-correlation functions [174]. The technical details surrounding ALEX/PIE are described in depth elsewhere [172], [174]. There have been several studies that have exploited these experimental approaches to investigate amyloid fibrils. For example, early oligomerisation kinetics and size distributions of a model amyloid forming protein α-spectrin SH3 have been characterised using PIE-FCS [175]. By co-aggregating differentially labelled α-spectrin SH3 and using PIE-FCS to observe and autocorrelate coincident events, the oligomeric distribution could be determined at each time point. This experiment was able to show the gradual appearance of a heterogeneous population of early-stage soluble oligomers, leading to the suggestion that some amyloid oligomers are unable to increase in size over time [175]. ALEX/PIE extend the capability of FCS and FCCS to probe the interaction between amyloid fibrils and the proteins they interact with, as well as the complex early stages of oligomerisation that have proven otherwise difficult to accurately characterise.

FRET

Fluorescence resonance energy transfer (FRET) is a photophysical phenomenon wherein energy is transferred from an excited-state donor fluorophore to an acceptor fluorophore by non-radiative dipole–dipole coupling (Fig. 4C), an effect that is inversely proportional to the distance between the donor and the acceptor (~10–100 Å, depending on the FRET pair used, referred to as the Förster distance) [176]. A large number of fluorophore pairs that exhibit FRET have been characterised, and suitable pairs can be selected to label proteins of interest depending on the type of information to be gathered (Fig. 4C) [177]. For example, FRET pairs can be used to label a single protein in two locations, such that when the protein is correctly folded, the FRET efficiency is high due to the fluorophores being in close proximity: when the protein is unfolded the FRET efficiency is lower due to the fluorophores being further apart [178]. Alternatively, to investigate protein–protein interactions, two separate proteins of interest can be labelled, each with one member of a FRET pair. High FRET efficiency would be expected if the proteins interact (and hence the FRET pair are in close proximity), and changes in FRET can be used to follow the dynamics of binding and dissociation [178].

FRET is a widely used approach for studying single molecules as it can be observed using both TIRF and confocal microscopy approaches [179], [180]. TIRF microscopy allows FRET from a single molecule to be measured through time, and thus kinetic information pertaining to molecular associations (including on and off rates) or changes in protein conformation can be elucidated [181], [182]. Alternatively, when used in conjunction with a confocal approach, FRET can report on molecular diffusion in tandem with protein-interactions. Such an approach avoids problems with steric hindrance that can occur as a result of the surface immobilisation required for TIRF microscopy [183], [184].

In recent years, several studies have utilised FRET as a technique to characterise oligomeric α-syn [76], [77], [78]. Well-characterised FRET pairs (such as AlexaFluor488 and AlexaFluor555 fluorophores) can be conjugated in equimolar amounts to monomeric α-syn, and the FRET efficiency measured in order to deduce precisely how monomers interact with one another during the early stages of the aggregation process. In these cases, a higher FRET efficiency is indicative of a closer proximity between molecules, or an altered dipole orientation of the fluorophore indicating larger or more compact oligomers [46]. Using a confocal single-molecule FRET approach, two structurally distinct oligomeric species of α-syn have been detected (named Type A and Type B) one of which is more compact and cytotoxic (Type B) [78]. The interconversion between these two species has been kinetically defined by measuring the transitions between FRET states, and this transition is the rate-limiting step in the formation of mature fibrils. The formation of mature fibrils following the transition to Type B is rapid, whilst the transition from Type A to Type B is slow [78]. This transition process may be a cyto-protective mechanism: it results in the burial of the exposed hydrophobic (and thus damaging) residues present in Type B oligomers. The proportion of each morphologically distinct oligomeric species was determined for a Parkinson’s disease-associated mutant form of α-syn [78]. The class of oligomer rapidly formed by A53T α-syn was the Type B oligomer, consistent with previous studies utilising ensemble-based approaches that have shown more rapid oligomerisation for this mutant [43]. In contrast, Type A oligomers predominated during the early stages of the aggregation of A30P mutant of α-syn, consistent with the finding that fibril formation is slower for the A30P mutant of α-syn [43], [185]. Thus, by using a confocal microscopy single-molecule FRET based approach, Tosatto et al. (2015) rationalised the kinetics of α-syn oligomerisation and provided unprecedented insight into the molecular mechanism that drives this disease-related process [78].

The functional yeast prion protein Ure2 also readily forms amyloid fibrils [186], and both oligomerisation and disaggregation of this protein have been investigated using combined TIRF and confocal single-molecule FRET approaches [46]. Conjugating donor and acceptor fluorophores in site-specific locations on the protein (chosen based on their expected Förster distance and dipole orientation upon β-sheet formation) showed that a conversion between oligomeric species is the rate-limiting step in the progression from soluble oligomers to amyloid fibrils. In particular, increasing FRET values over time suggested a more compact oligomer is present immediately prior to fibril elongation. This transition to a more compact state is invisible in ensemble biochemistry techniques as it occurs within the ‘lag’ phase (prior to ThT increasing in fluorescence during the exponential growth phase). The disaggregation of Ure2 [46] and α-syn [77] has also been examined using a single-molecule FRET approach. Oligomeric species released from mature fibrils of both proteins have FRET values that mirror those of species detected immediately prior to amyloid fibril elongation, suggesting that disaggregation results in on-pathway species that could propagate further fibril formation.

Studying the interaction of aggregates with other proteins using fluorescence-based single-molecule techniques

Single-molecule fluorescence methods can be used to detect protein–protein interactions involved in some of the most complex molecular processes that underpin life and disease states. This includes membrane-protein interactions involved in cell-signalling processes [187], [188], and DNA-replication machinery [189]. A key example of a cellular process that uniquely benefits from such approaches is the interplay between protein aggregates and molecular chaperones.

In order to avoid the formation of toxic protein aggregates, cells employ a number of protein quality control mechanisms, including molecular chaperones, which engage with destabilised proteins in order to prevent them from aggregating (for more detailed reviews on the mechanism of action of molecular chaperones, see [190], [191]). Moreover, molecular chaperones have been found to co-localise with amyloid fibrils within the proteinaceous inclusions in the neurons of patients with neurodegenerative diseases [192], [193], [194], [195], [196]. However, the interaction between molecular chaperones and amyloid fibril forming proteins can be difficult to characterise using bulk biochemical techniques owing to the highly heterogeneous and dynamic nature of both systems [190]. As a consequence, single-molecule fluorescence approaches have been invaluable in elucidating the interactions between molecular chaperones and amyloid fibril forming proteins.

The aggregation of the protein tau is associated with a number of neurodegenerative diseases, most notably Alzheimer’s disease. By conjugating a donor fluorophore (AlexaFluor488) to tau, and an acceptor (AlexaFluor647) to the ATP-dependent chaperone Hsp70, interactions between these two species could be detected by FRET [197]. This approach enabled the size of both Hsp70 and the tau aggregates in samples to be determined and the dissociation rate of Hsp70 from different tau species to be determined. In doing so, this study established that Hsp70 binds to and then rapidly dissociates from tau monomers, yet forms increasingly stable binding interactions with oligomers and fibrils of tau (as a function of aggregate size).

A study by Cox et al. (2018) examined the binding of small heat shock proteins (sHsp) to α-syn fibrils by using the widely available AIE dyes ThT and Nile red, as well as AlexaFluor647 conjugated to the ATP-independent sHsp Hsp27 [112]. Using TIRF microscopy with dual excitation, individual α-syn fibrils were visualised by the emission of the ThT dye (at 480 nm) and the localisation of Hsp27 on the fibrils established by the emission at 670 nm. This work demonstrated that Hsp27 forms stable complexes with α-syn fibrils. Moreover, by using Nile Red in place of ThT for the imaging of the fibrils, it was demonstrated that the fibrils bound by Hsp27 had reduced surface hydrophobicity and this correlated with reduced cytotoxicity of the Hsp27-coated fibrils [112]. Whilst single-molecule fluorescence microscopy has enhanced the ability to observe stable interactions between sHsps and amyloid fibrils, there is still very little known about the specific mechanisms responsible for sHsp chaperone function. Moreover, there has been wide speculation on the precise role of the various oligomeric forms of sHsps when it comes to interacting with fibrils [198], [199], [200]. Recently, this type of mechanistic detail has been investigated using single-molecule photobleaching in the context of a protein that forms amorphous aggregates [201]; a similar experimental approach could be employed to study the interaction of sHsps with amyloid fibrils.

The extracellular chaperone clusterin can bind to and reduce the toxicity of oligomeric forms of α-syn [202]. By labelling α-syn and clusterin with AlexaFluor488 and AlexaFluor647 (respectively), Whiten et al. (2018) followed the interaction of these two proteins by TCCD under conditions that promote α-syn aggregation. Supported by ensemble-based enzyme-linked immunosorbent assays, this single-molecule work determined the apparent stoichiometry of clusterin bound to α-syn over time by comparing the relative fluorescence of coincident bursts within the confocal volume. During the aggregation of α-syn, the amount of clusterin that was bound to α-syn reduced as the size of the α-syn oligomers increased. From this it was concluded that clusterin binds to hydrophobic regions on the surface of α-syn fibrils: as the aggregation of α-syn proceeds the surface hydrophobicity of fibrils decreases and this leads to a decrease in the amount of bound clusterin.

Apolipoprotein E (ApoE) has been found within Aβ plaques in the brains of patients with AD [203], [204], and mutations in the gene encoding ApoE (the polymorphic ε2, ε3, and ε4 APOE alleles) increase the severity and likelihood of both early and late onset forms of AD (in particular the ε4 allele (ApoE4), which is the strongest risk factor for AD) [205] (reviewed in depth elsewhere, see [159]). Both ALEX-FCCS and FCS have been used to detect complexes formed between Aβ(1-40), ApoE3 and ApoE4 [206]. By labelling the ApoE mutants and Aβ(1-40) with AlexaFluor488 and AlexaFluor647 respectively, and analysing their diffusion on a confocal microscope using alternating laser excitation to reduce fluorophore cross-talk [174], fluorescence fluctuations were cross-correlated and analysed for the presence of complexes. An association between the two proteins was established using this method, suggesting that there is an interaction between Aβ(1-40) and ApoE [206]. Furthermore, the auto-correlation of Aβ during amyloid formation showed the species formed had a smaller hydrodynamic radius in the presence of both ApoE mutants compared to when ApoE was not present. This suggests that by forming complexes with Aβ, ApoE reduces the size of the Aβ aggregates. Interestingly, this effect on the size of the Aβ aggregates was reduced in the presence of ApoE4 compared to ApoE3, suggesting that the increased risk of disease associated with this mutation is due to a decrease in the capacity of ApoE to bind Aβ(1-40) and inhibit amyloid fibril formation. This result provides compelling evidence as to how mutations in ApoE may increase the risk and severity of AD. Moreover, the use of ALEX-FCCS and FCS in this work highlights how single-molecule fluorescence techniques can be used to uncover molecular mechanisms that relate to the onset and progression of diseases associated with amyloid fibril formation. Elucidating the details of these dynamic biomolecular associations is not possible using post-mortem tissues from patients [207], [208].

The disaggregation of tau [69] and α-syn [209], [210] fibrils by the Hsp70 system has been previously documented in vitro; however, a detailed mechanistic model for this process remained to be established. Recently, the disaggregation of α-syn by the Hsp70 disaggregase system has been studied using ensemble-averaged FRET, fluorescence anisotropy, and kinetic monte-carlo simulations [68]. This work proposed a model in which the Hsp40 co-chaperone selectively binds oligomeric forms of α-syn in order to sequester Hsp70 onto amyloid structures, in turn facilitating entropic pulling via molecular crowding of the chaperones [68]. This mechanism was established by labelling equimolar amounts of Hsp70 molecules with a FRET pair such that, if crowding of Hsps70 molecules on the α-syn fibril occurred, the FRET efficiency would be high. In contrast, if the Hsp70 molecules were spread out along the fibril, the FRET efficiency would be low. Furthermore, a more dramatic increase in FRET efficiency in the presence of the NEF Hsp110 demonstrated that Hsp110 acts to increase the propensity for molecular crowding of Hsp70, enhancing entropic pulling. This mechanistic model provides evidence for a previously proposed theory that entropic pulling is responsible for the depolymerisation or fragmentation of amyloid fibrils [211], [212]. Conclusive evidence for this mechanistic model could be provided by using TIRF-based experiments, including fluorophore photobleaching to directly observe and quantify molecular crowding, the position on the fibril at which this occurs, and the binding and dissociation rates of Hsp70/40 and NEF Hsp110. Moreover, the discovery of new AIE dyes with large Stokes shifts, such as ASCP [88], [108] provide the possibility to visualise fibrils and proteins that interact with them using only one excitation source [108]. More complex three-colour fluorescence experiments would enable more components to be observed simultaneously [213], [214] and thus be invaluable for the study of how chaperone machines, such as how the Hsp70/Hsp40/Hsp110 disaggregase system [66], [75], interact with amyloid fibrils.

Summary and outlook

This review has focussed on the use of single-molecule fluorescence-based approaches for the study of the molecular details underpinning amyloid fibril formation and their interaction with other proteins, in particular molecular chaperones. Understanding mechanistic details of the oligomerisation process, as well as the protein–protein interactions that interfere with it, is an important step towards the development of novel drugs to treat or even prevent diseases such as Alzheimer’s and Parkinson’s disease. As a result of being highly heterogeneous, some of the mechanistic details of amyloid fibril formation are difficult to ascertain using ensemble-based techniques. Single-molecule fluorescence methods, such as TCCD, FCS and FRET, have already provided insight into these dynamic and heterogeneous processes. Recent work on the mechanism by which molecular chaperones bind to fibrils [112], [119], [197], [206], and the species released during the disaggregation of fibrils [74], [149] are excellent examples of this. However, there is still considerable scope to apply and extend these single-molecule methods for the study of fibrils and the manner in which they interact with other proteins.

Finally, it is important to recognise that these single-molecule fluorescence-based methods are most powerful when used in combination with other complimentary approaches. For example, techniques such as solid-state or hydrogen–deuterium exchange NMR or electron microscopy, can provide information about the structural polymorphisms of amyloid fibrils, and the impact of fluorescent tagging on these structures. It will be through the use of highly integrated approaches such as these that we will gain further fundamental knowledge that is critical to our understanding of how neurodegenerative diseases arise, and how we may interfere with this process in order to treat or prevent them.

CRediT authorship contribution statement

Lauren J. Rice: Conceptualization, Investigation, Writing – original draft, Visualization, Writing - review & editing. Heath Ecroyd: Conceptualization, Project administration, Writing - review & editing. Antoine M. van Oijen: Conceptualization, Project administration, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.