Molecular interactions of amyloid nanofibrils with biological aggregation modifiers: implications for cytotoxicity mechanisms and biomaterial design

Abstract

Amyloid nanofibrils are ubiquitous biological protein fibrous aggregates, with a wide range of either toxic or beneficial activities that are relevant to human disease and normal biology. Protein amyloid fibrillization occurs via nucleated polymerization, through non-covalent interactions. As such, protein nanofibril formation is based on a complex interplay between kinetic and thermodynamic factors. The process entails metastable oligomeric species and a highly thermodynamically favoured end state. The kinetics, and the reaction pathway itself, can be influenced by third party moieties, either molecules or surfaces. Specifically, in the biological context, different classes of biomolecules are known to act as catalysts, inhibitors or modifiers of the generic protein fibrillization process. The biological aggregation modifiers reviewed here include lipid membranes of varying composition, glycosaminoglycans and metal ions, with a final word on xenobiotic compounds. The corresponding molecular interactions are critically analysed and placed in the context of the mechanisms of cytotoxicity of the amyloids involved in diverse pathologies and the non-toxicity of functional amyloids (at least towards their biological host). Finally, the utilization of this knowledge towards the design of bio-inspired and biocompatible nanomaterials is explored.

1. Background on amyloid nanofibrils: structural features and biological relevance

Protein aggregation into nanoscale amyloid fibrils is a widespread phenomenon in biology. While all amyloids share common structural features, their biological functionality ranges from highly toxic to safe and functional. The amyloid structure is recognized as a generic fold and as an intrinsic property of the polypeptide chain under suitable conditions [1,2]. In addition, the amyloid structure is believed to be one of the first functional protein folds evolved in living cells, and potentially in the prebiotic world [3].

1.1. Common structural features

Amyloid nanofibrils are biological peptide/protein fibrous structures that are composed of extended β-sheet hydrogen bond networks running along the fibril axis, orthogonally to the peptide/protein backbone array [4,5]. The monomers or soluble peptide/protein molecules self-assemble into oligomers, which give rise to protofibrils that further pack to form mature amyloid fibres [5,6]. Amyloid nanofibrils show characteristic features in electron microscopy, X-ray fibre diffraction and specific dye binding [5,7]. Typically, generic amyloid fibrils are straight, unbranched, exhibit a diameter of 6–12 nm and are up to a few microns long (figure 1).

Figure 1.

Electron micrograph of somatostatin functional amyloid nanofibrils spontaneously formed in water within a week, peptide concentration 5% w/w (D Dharmadana, C Valery 2015, unpublished data). The preformed fibrils were stained with 1% uranyl acetate, using a previously published protocol [8]. The peptide is a human neuropeptide hormone, reported to undergo fibrillization under various conditions for storage purposes within brain secretory granules [8,9].

Typical X-ray diffraction patterns of the cross-β structure for aligned fibrils exhibit a meridional reflection at 4.8 Å, which corresponds to the repetitive distance between β strands, and an equatorial reflection at 10–11 Å, which arises from the distance between β-sheet networks (figure 2). Detailed structural investigations by X-ray crystallography on amyloid nanofibril single crystals have so far revealed eight different classes of molecular packing in the cross-β organization, as a function of the peptide/protein side chains [10]. Solid-state nuclear magnetic resonance structure resolution provided a molecular structural basis for explaining the amyloid width polymorphism observed in many systems [11].

Figure 2.

The cross-β-sheet motif and corresponding X-ray diffraction pattern. (Reproduced with permission from Riek & Eisenberg [10] (Copyright © 2016 Nature publishing group).) (Online version in colour.)

The role of amyloid protein fibrils in neurodegenerative disease has been the subject of significant research focus. However, more recently, functional amyloids have also been discovered, which possess almost identical morphologies to toxic amyloids, yet are non-toxic with essential biological functions (tables 1 and 2).

Table 1.

Examples of functional amyloids characterized in prokaryotes and eukaryotes.

amyloid-forming protein	associated biological function	organism
curli	biofilm formation	Escherichia coli [12]
microcin E492	channel-forming bacteriocin	Enterobacteriaceae [13]
chaplins	assisting spore dispersal and soil colonization	Streptomyces coelicolor [14]
harpins	plant invasion and pathology	Xanthomonas species [15]
Ure2p	regulation of nitrogen catabolism	Saccharomyces cerevisiae [16]
hydrophobins	formation of hydrophobic coating on surfaces by filamentous fungi	Schizophyllum commune [17]
chorion protein	protecting the oocyte and developing embryo from environmental hazards	silk worm [18]
spidroin protein	intermediate in the biosynthesis of spider silk	spider [19]
Pmel17	assisting with melanin production	mammalians [20]
42 brain hormones	numerous regulatory functions; density gain for storage within secretory granules	humans [8,9]

Table 2.

Misfolding diseases due to protein aggregation into amyloid nanofibrils and corresponding protein/peptide. (Adapted from [21].)

protein/peptide-forming amyloid	protein misfolding disease
Aβ peptide	Alzheimer's disease
α-synuclein	Parkinson's disease
islet amyloid polyprotein (IAPP/amylin)	diabetes mellitus type 2
huntingtin	Huntington's disease
monoclonal immunoglobulin light/heavy chain	immunoglobulin light/heavy-chain amyloidosis
prion protein	spongiform encephalopathy Creutzfeldt–Jakob disease
plasma transthyretin	senile systemic amyloidosis
calcitonin	medullary carcinoma of the thyroid
prolactin	prolactinomas
apolipoprotein AI	atherosclerosis

1.2. Functional amyloids

In the past decade, a significant number of natural peptides and proteins have been shown to self-assemble into non-toxic amyloid-like nanofibrils both in vivo and in vitro (table 1). Physiological functions have been associated with these so-called functional amyloids, built from the self-assembly of natural peptides/proteins in their native conformation, under physiological conditions. Functional amyloids have been characterized in a wide range of organisms and are extensively reviewed elsewhere [22–24] (table 1).

Non-mammalian examples include the Escherichia coli curli protein nanofibrils associated with biofilm formation [12], silk worm chorion protein amyloids that protect the oocyte and developing embryo from environmental hazards [18] and the spider spidroin protein nanofibrils produced in the biosynthesis process of spider web silk fibres [19]. Mammalian and human examples include nanofibrils assembled from the Pmel17 protein, which assists in melanin production [20], and a significant number of brain peptide and protein hormones that self-assemble for storage purposes within intracellular secretory granules [8,9]. All these examples exhibit the structural features of typical amyloid systems, without being cytotoxic to their host.

The lack of cytotoxicity of functional amyloids is currently thought to be due to their ability to reversibly self-assemble [9], due to the relative size and surface hydrophobicity of the oligomers formed during the fibrillization process [25] and/or to the biological inhibition of the fibrillization process by protein chaperones [26].

1.3. Disease-associated amyloids

Over 20 disease conditions are associated with amyloid aggregates and nanofibril formation, including Alzheimer's disease (AD) and Parkinson's disease (PD) [27,28]. Disease conditions involving protein aggregation and amyloid formation are known as protein misfolding diseases. In each pathology, a different protein (or proteins) misfolds and forms amyloid nanofibrils that then go on to form the major component of micrometre-sized insoluble deposits on the affected organs (table 2) [21].

For instance, the amyloid β peptide (Aβ) (produced by cleavage of the amyloid precursor protein) and Tau protein both undergo conformational changes to form water-insoluble nanofibrils in the brain in AD. These amyloid nanofibrils then form extracellular neuronal plaques (Aβ) and intracellular neurofibrillary tangles (Tau), which are the major pathological hallmark of AD [21,29]. In PD the protein α-synuclein undergoes a refolding event promoting fibrillization, eventually leading to the deposition of Lewy bodies in the brain of affected individuals. Similarly, a number of degenerative conditions and brain function alterations result from the toxicity of other amyloid formation processes [21].

The molecular details of the mechanisms leading to cell or tissue death in amyloid diseases remain unknown. However, amyloid toxicity hypotheses involving misfolded protein oligomers formed in the early steps of the fibrillization process are currently supported by a number of in vitro structural studies, as reviewed in the following sections.

2. Amyloid fibrillization process: generic features and molecular mechanisms

The conversion of soluble protein species into amyloid aggregates has been widely studied over the past decade. Amyloid fibrillization can be described by three characteristic kinetic phases: a lag phase, a growth phase and a final saturation phase. Nucleated polymerization is recognized as the underlying mechanism [30–32].

The conversion of soluble protein species into β-sheet-based nuclei takes place during the lag phase via multiple parallel processes, generally involving a large variety of metastable oligomers [33]. Some nuclei then elongate and proliferate until reaching a detectable aggregate concentration, hence terminating the experimentally observed lag phase. Aggregate elongation and replication further continues until saturation. A significant number of different amyloid systems have been shown to involve secondary nucleation mechanisms that favour the elongation and replication steps: nanofibril surfaces can catalyse the formation of new aggregates, or nanofibrils can undergo fragmentation [30] (figure 3).

Figure 3.

Species involved in the amyloid fibrillization process (a) and a generic kinetic scheme of aggregation (b).

The amyloid fibrillization process exhibits high complexity in the very first steps of conversion, from protein monomers to nuclei that can elongate. A range of mechanisms have been proposed for these early stages, including nucleated conformational conversion [34]. Protein soluble monomers have been proposed to influence the conversion of other native monomers into prefibrillar nuclei, which then eventually elongate into amyloid fibrils [35,36]. The kinetic constants within the overall fibrillization process are highly influenced by the concentration of the species involved. An increase in concentration of the soluble protein species, or the addition of preformed nuclei or fibrils to a system, has been shown to reduce the lag phase and to promote fibrillization [37–39]. From a thermodynamic point of view, nuclei formation is unfavourable, while the elongation step represents a loss in Gibbs free energy. Indeed, the amyloid fibril fold is one of the most thermodynamically stable protein folding states [27,31,40–42] (figure 4).

Figure 4.

Energy landscape of protein folding and aggregation. (Reproduced with permission from Hartl & Hayer-Hartl [41] (Copyright © 2016, Nature publishing group).)

A large variety of metastable oligomers are generally formed in the early steps of the fibrillization process. These can differ in their conformation, morphology (globular or annular), seeding capacity and cytotoxicity [33]. Experimental studies in vivo support that protein oligomers are the major toxic species, rather than the mature fibrils [43–45]. The latest structure–activity rationales rely on subtle physicochemical parameters to rationalize the cellular toxicity of oligomers, such as surface hydrophobicity and size. It was recently proposed that small soluble oligomers are less damaging to cells than larger oligomers exhibiting a hydrophobic surface [25]. However, the molecular mechanisms of cytotoxicity appear to be complex and may not involve a single toxic species. Furthermore, interactions with biomolecules and biosurfaces may also play a role in the cytotoxic mechanism. These interactions, specifically focused on lipid membranes, glycosaminoglycans (GAGs) and biometal ions, will be critically reviewed in the following sections.

3. Interactions of amyloids with cell/lipid membranes

The interaction of amyloid-forming peptides with cellular and model lipid membranes is twofold. On the one hand, interaction with the lipid bilayer structure modulates amyloid fibrillization, promoting rather than inhibiting fibril growth in the majority of cases [46]. Conversely, the membrane structure can be significantly disrupted by interaction with the peptide, resulting in toxicity [47] (figure 5).

Figure 5.

Three proposed mechanisms of cell membrane disruption caused by amyloid oligomers or fibrils. (i) The channel hypothesis, whereby nanoscale annular amyloid oligomers insert into the bilayer forming ‘ion channels’ across the membrane. (ii) Less ordered insertion of amyloids into the bilayer resulting in reduced lipid mobility, membrane thinning or non-specific deregulation of membrane homeostasis. (iii) Amyloid–lipid co-aggregation whereby the process of amyloid aggregation results in the extraction of lipids from the bilayer and eventually a loss of membrane fidelity.

3.1. Modulation of fibrillization by lipid composition

The interactions between amyloid-forming fibrils and individual lipids have been mainly characterized for toxic peptides, with a predominant focus on the Aβ peptide [47], α-synuclein and islet amyloid precursor protein (IAPP). Rather less information exists on the interaction between functional amyloid-forming peptides and lipid membranes of specific composition. The roles of the lipid membrane in fibrillization may include increasing the local concentration of bound peptide, and unfolding and orienting the peptide in such a way as to facilitate the peptide–peptide contacts necessary for fibrillization to occur [47]. Specific interactions have been noted for a range of lipid types including anionic lipids, cholesterol, gangliosides and phosphatidylinositol. In the majority of cases lipid interactions appear to promote rather than inhibit fibril growth.

3.1.1. Anionic lipids

The majority of amyloid-forming peptides (including Aβ and IAPP) are cationic under physiological conditions. For α-synuclein, which has a pI of 4.5 and is basic at neutral pH, the N-terminal membrane-binding region is cationic. Electrostatic interactions with lipid membranes containing anionic lipids can therefore facilitate the initial binding of the protein or peptide to the membrane. Aβ(1−40) has been observed to insert spontaneously into anionic (1,2-dipalmitoyl phosphatidylglycerol) DPPG membranes, with no interaction between this peptide and zwitterionic (1,2-dipalmitoyl phosphatidylcholine) DPPC membranes observed [48]. The interaction was reduced in the presence of 0.1 M NaCl solutions, confirming that the interaction is mainly electrostatic rather than hydrophobic. Circular dichroism spectroscopy indicated that binding of the peptide to the anionic lipid bilayer promoted a transition from a random coil to a β-sheet conformation for Aβ [49]. Molecular dynamics simulations have confirmed that peptide–peptide interactions promote oligomerization for anionic membranes [50].

3.1.2. Cholesterol

The role of cholesterol in fibrillization appears to depend strongly on its molar composition within the bilayer. Cholesterol has been shown to both promote [51] and inhibit [52] the interaction between Aβ(1−40) and model lipid bilayers. This may be linked to the role of cholesterol in modulating membrane fluidity as the fluidity of the bilayer has been shown to impact its ability to promote fibrillization [53]. Low concentrations of cholesterol in the bilayer are known to increase the rigidity of the bilayer, while the membrane may actually become more fluid at high cholesterol concentrations.

3.1.3. Gangliosides and lipid rafts

Gangliosides, containing an anionic sialic acid headgroup, are a type of glycolipid. They are believed to have a role in the formation of lipid rafts, rigid membrane micro-domains formed by the clustering of cholesterol, sphingolipids and glycolipids, and are predominantly found in neuronal cells [53]. As for cholesterol, they have been shown to promote fibrillization for Aβ [54] and IAPP [55] and to inhibit fibrillization for Aβ, potentially linked to the peptide–gangliodside ratio. A correlation between AD and low levels of gangliosides in the cell membrane has also been observed [56]. Aβ and IAPP have been shown to preferentially accumulate on rigid lipid rafts; conversely, α-synuclein has been seen to accumulate on more disordered lipid membranes [57].

3.2. Mechanism of toxicity

It has been reported that aggregates of Aβ and other disease-associated amyloid-forming proteins are toxic to several types of cells [58–60]. Despite extensive research, it is still unclear which aggregated species (e.g. oligomer, protofilament, mature fibrils) is responsible for toxicity. A growing body of evidence suggests that oligomers are responsible for toxicity due to damage inflicted on the cell membrane, disrupting homeostasis and triggering a cascade of events eventually leading to cell death [61–63]. However, evidence for the toxicity of Lewy bodies (deposits of mature amyloid fibrils and other biomolecules found in PD patients) [64] and a number of studies showing that membrane damage occurs in the presence of other assemblies suggest that the toxic species cannot be definitively assigned to oligomers alone [65–72].

The molecular mechanism(s) responsible for membrane damage is even less clear. It has been shown that oligomers from a number of amyloidogenic proteins including Aβ [73,74], α-synuclein [75,76] and IAPP [77] can form nanoscale pores in reductionist model membrane systems such as supported lipid bilayers (SLBs) and unilamellar vesicles. These results lead to the compelling ‘channel hypothesis’ whereby annular oligomers form pores in the cell membrane disrupting the regulation of calcium (and other ions) leading to cell death [78]. Supporting the channel hypothesis, Kawahara et al. [79] showed that Aβ can form annular protofilaments in excised membrane patches from hypothalamic neurons. Demuro et al. [80] showed that calcium deregulation and membrane disruption occurred in SH-SY5Y cells for a range of amyloid oligomers, and similar evidence of in vitro calcium disruption has been seen for Aβ [81,82] and α-synuclein [83,84]. However, amyloid ‘ion channels’ do not represent the only possible explanation for deregulated calcium influx into cells [85]. Indeed, in the 30 years or so since the proposal of the channel hypothesis, there has been little direct evidence for the occurrence of this mechanism in living cells.

A number of alternative hypotheses that might result in a similar loss of membrane homeostasis have been proposed. For instance Ouberai et al. [86] demonstrated that α-synuclein binds to the packing defects in a lipid bilayer, causing membrane remodelling. Excessive membrane tubulation [87,88] or membrane thinning [89] induced by amyloid proteins have also been proposed as possible mechanisms of toxicity. One particularly intriguing possibility is that of amyloid–lipid co-aggregation, whereby growing oligomers or fibrils extract lipids from the membrane resulting in membrane damage and eventually a loss of fidelity. Mechanisms involving lipid extraction and co-aggregation have been proposed for α-synuclein [66,69–71,90], amylin [67,68,91] and β₂m [92]. Furthermore, Lashuel and co-workers have shown that, for both α-synuclein [65] and Aβ [93], monomers, oligomers or fibres alone are not highly toxic to neurons, but when oligomers or fibres are mixed with monomers allowing continued aggregation to occur their toxicity increases dramatically. A number of in vivo studies also support a co-aggregation hypothesis: Lewy bodies extracted from PD sufferers post mortem were found to contain high levels of lipids [94], and lipid-bound α-synuclein oligomers were found in the brains of PD mouse models and humans suffering from Lewy body dementia [95]. Continuing aggregation also has been shown to be a requirement for toxicity in systematic amyloidosis [96] and prion diseases [97], suggesting that the ‘co-aggregation hypothesis’ may represent a universal mechanism of toxicity across many amyloid diseases.

The search for a definitive toxic mechanism whereby amyloid aggregates disrupt cell membranes and trigger cell death remains elusive, with seemingly contradictory evidence supporting a number of hypotheses. This may reflect, in part, the reliance of much of this research on highly reductionist SLB or vesicular systems, which bear little resemblance to the cell membrane. Future efforts should be directed to increasing the complexity of the model membranes used to better mimic neuronal cell membranes.

One thing that seems clear is that a single definitive mechanism of toxicity is unlikely. Neurotoxicity and cell death, even within the same neurodegenerative disease, most probably occur via a number of different molecular mechanisms. Thus, the task of identifying these mechanisms and designing effective therapeutic strategies is daunting. However, recent breakthroughs such as the development of the antibody aducanumab to reduce Aβ plaques in AD provide reasons to be hopeful that curative therapeutic strategies may be on the horizon [97].

While a large proportion of research investigating the interactions of amyloids with lipid membranes of cells has focused on discovering mechanisms of toxicity, there has been some research into the interactions of functional amyloids with cells and also synthetic amyloid-like fibres which have the potential to act as biomimetic materials that promote cell growth. In the case of functional amyloid cytotoxicity, Singh & Maji [98] showed that tachykynin neuropeptides are non-toxic to neuroblastoma cells. A study conducted on the cytotoxicity of HypF-N protein oligomers proposed a correlation between oligomer hydrophobicity and size to predict oligomer toxicity [25]. According to that study, toxic oligomers have high hydrophobicity and small size. Substrate-bound networks of amyloids formed from non-toxic food-grade proteins such as hen egg white lysozyme or whey protein have been shown to form fibres with little to no toxicity, and have beneficial effects on cultured cells [99,100]. This is thought to be due to the biomimetic topography of these fibrils, which closely resembles that of the extracellular matrix (ECM) [101]. Synthetic amyloids formed from short peptide sequences identified from amyloidogenic proteins are also attractive as biomaterials, as their sequences can be altered to display sequences (i.e. RGD) from the ECM that promote cell adhesion and growth [102–105]. At high concentrations, many amyloid fibril networks form highly hydrated three-dimensional (3D) networks, generally referred to as hydrogels. Amyloid hydrogels can encapsulate many therapeutic and biological molecules and make attractive materials for 3D cell culture. Such hydrogels have found applications in the study and control of cellular responses in highly biomimetic 3D environments [104,105], as implantable materials for regenerative medicine [105] and as micro (or nano) carriers for drug delivery applications [106,107]. For example, the Maji group fabricated hydrogels from fibrils assembled from short peptide sequences identified in Aβ [104] and α-synculein [105]. The stiffness of the Aβ hydrogels could be tuned, enabling control of mesenchymal stem cell (MSC) differentiation [104]. The α-synculein hydrogels were able to promote the neuronal differentiation of encapsulated MSCs both in vitro and after implantation into the brains of mice [105]. Studies such as this clearly show that amyloid-based 3D scaffolds hold great promise for regenerative medical and tissue engineering applications.

4. Glycosaminoglycans as catalytic electrostatic scaffolds

Glycosaminoglycans (GAGs) are biological polysaccharides of 10–100 kDa and major components of the ECM in multicellular organisms. GAGs consist of repeating disaccharide units—a uronic acid unit and an amino sugar—with varying degrees of O- and N-sulfation, making most GAGs highly anionic polyelectrolytes [108]. In the biological context, GAGs are known to interact with cell surfaces and proteins, and to eventually form GAG–protein complexes called proteoglycans. Initially thought of as an inert scaffold, the ECM is now known to exhibit biological activity and GAGs may modulate a wide range of biological functions [109,110]. Heparin/heparan sulfate and corresponding proteoglycans have been the main focus for studies of GAG–amyloid interactions. This choice is motivated by the presence of heparin and associated species as the main GAG component within the ECM and within diverse disease-related amyloid plaques [111–113]. For example, heparin/heparan sulfate has been identified within the amyloid plaques of AD and PD, type 2 diabetes, light chain amyloidosis and prion-related diseases [114–116]. GAG–amyloid interactions have been studied both in vivo and in vitro, for both toxic and functional amyloids, and within the context of biomaterial design, such as amyloid-based nanoscaffolds for 3D cell culture [117].

Heparin induces and/or enhances amyloid fibrillization in vitro for a range of biological peptide/protein systems [118], including the disease-related Aβ peptides [119–121] and Tau protein [122,123], α-synuclein [113], gelsolin [124,125], IAPP [126], and a range of neuropeptide hormones and neurotransmitters [9]. From a kinetic point of view, GAGs, and specifically heparin, significantly reduce the lag phase, increase the kinetics of the growth phase and/or induce fibrillization, as reported for all the amyloidogenic systems mentioned above. Cohlberg et al. [113] showed that heparin has a high affinity for α-synuclein and not only accelerates fibrillization but also increases the fibril yield. Maji et al. [9] reported that a significant number of neuropeptide hormones require heparin to induce functional fibrillogenesis in vitro. Adamcik et al. [122] showed that heparin strongly modulates the assembly of the R3 peptide sequence (the microtubule-binding sequence from Tau), promoting amyloid formation. In the absence of heparin, R3 forms fibrils with a strong tendency to laterally associate, with fibrils consisting of up to 45 protofilaments observed. This work highlights the strong effect of heparin, not just on the extent of assembly but also on the dominating intermolecular interactions and the morphology of the end structures.

The promoting effect of GAGs on fibrillization is mainly due to electrostatic interactions. These occur between the anionic GAG polyelectrolytes and the cationic side chains of the amyloidogenic proteins. However, van der Waals forces, hydrogen bonds and hydrophobic interactions between the protein and the GAG carbohydrate backbone may also play a role [127]. Supporting the main electrostatic nature of the GAG–amyloid species interactions, histidine protonation was reported to impact the binding affinity of IAPP for heparin, and, concomitantly, the promoter effect of heparin on IAPP fibrillization [128]. In another study, acetylation of heparin sulfate groups (thus neutralizing) was shown to result in the loss of the heparin inducer effect on Aβ1–40 fibrillization [119]. Further, electrostatic screening by calcium or magnesium ions introduced in the bulk solution resulted in a similar loss of the heparin promoter effect, as reported for muscle acylphosphatase and human lysozyme fibrillization [129]. From studies on the wild-type and mutants of the amyloid light chain protein, Blancas-Mejía et al. [130] showed that the GAGs’ effect depends on their degree of sulfation, and hence charge. Specifically, heparin was reported to promote the soluble protein conversion into amyloid nuclei that can further elongate, while chondroitin, a less sulfated GAG, was proposed to kinetically trap unfolded intermediates, resulting in the observed fibrillization inhibition. Although electrostatic interactions are strong molecular interactions, the GAG–amyloid species pairing is subject to a relatively subtle balance, with potential opposite consequences on the end structures, but still within the fibrillization pathway.

Scaffolding-based mechanisms are currently accepted for the GAG–amyloid species interactions, in which the GAG polyelectrolyte chains direct nuclei binding and amyloid elongation by electrostatic interactions, via their regularly spaced sulfate charges [124,131] (figure 6). These interactions take place from the very early stages of the fibrillization process, with however some discrepancies from one system to another, depending on which protein species can efficiently bind to GAGs. Heparin was reported to only bind to gelsolin cross-β-sheet oligomers and subsequent elongated oligomers, probably due to different side chain exposures for the other protein species [124]. Similarly, GAGs were shown to bind to transthyretin preformed misfolded oligomers and to promote their elongation, but not to bind to the initial native or unfolded species [132]. On the contrary, heparan sulfate was shown to promote the early conversion of incompletely processed proIAPP into misfolded intermediates that further elongate into amyloid fibrils [126]. GAG effects have hence to be considered on a case-by-case basis, although generic mechanistic principles apply.

Figure 6.

Proposed mechanism of heparin-mediated enhancement of protein self-assembly into amyloid fibrils. Some discrepancies exist from one protein system to another, depending on which protein species can efficiently bind to heparin/GAGs (see text).

In addition to binding amyloid species and acting as scaffolds for conformational conversion or the elongation of supramolecular networks, GAGs have been shown to reduce the cytotoxicity of a significant number of amyloid systems. Three main mechanisms have been proposed. GAGs can reduce toxicity by:

• (i) Promoting the conversion of toxic oligomers into less toxic aggregates/fibrils, such as for IAPP [128]. The gradual loss of aggregate solubility with size increase may play a role.

• (ii) Hindering the formation of toxic species by trapping less toxic intermediates, such as for the prion protein [133]. This is the other side of the toxicity spectrum, when compared with the previous mechanism. Here, small soluble intermediates, either unfolded monomers or small oligomers, are not yet toxic.

• (iii) Inhibiting the interactions of toxic species with cells, by binding to the cell surface, such as for the HypF-N protein. HypF-N protein, in a native, misfolded or oligomerized state, does not bind to GAGs [134]. The toxicity loss is due to the capacity of GAGs to screen the cell surface–ECM interactions.

Amyloid cytotoxicity in the presence of GAGs can hence be variable, given the range of mechanisms that can be involved. Taken together, these works reveal that generic assumptions are difficult to make on the effects of GAGs on amyloid fibrillization and cytotoxicity. Rather, GAG–amyloid interactions should be considered case by case, as a function of the type of GAG, its charge, the amyloid system and the binding affinity of the whole range of protein amyloid species. When GAGs interact with amyloid systems, they are likely to modify the kinetics of the fibrillization process, the end structure morphology and the cytotoxicity. GAGs can act as promoters of the complex fibrillization pathway, via electrostatic scaffolding. Discrepancies of the interacting protein species have been noted between amyloid systems. Eventually, GAGs trap protein intermediates, then inhibiting fibrillization from proceeding further.

Given their potential interactions with peptide/protein self-assembling systems, specifically amyloid-like synthetic assemblies, GAGs are currently considered as components of synthetic bioscaffolds in bionanotechology. For instance, GAGs have been used to promote the assembly of de novo designed peptides into nanofibrils, with applications as bioscaffolds for blood vessel growth [135] or for 3D cell culture [117]. GAGs can hence be useful in the domain of amyloid-like biomaterials, to help in controlling the kinetics, morphology, cytotoxicity or activity of synthetic peptide/protein nanostructures.

5. Interactions of metal ions with amyloid systems

There is growing evidence supporting the binding of metal ions to amyloid species, specifically in the context of misfolding diseases. Aberrant metal ion homeostasis is characterized in the brain of patients for several amyloid conditions. High concentrations of Cu(II), Zn(II) and Fe(III) have been identified in the core and rim of AD senile plaques, with metal ions clearly bound to Aβ peptides [136,137]. Binding of zinc and copper ions to Aβ peptides has been proposed to induce fibrillization within synaptic clefts, where some neuron subclasses eventually release high concentrations of metal ions [138,139], specifically in the hippocampus, a region of the brain involved in memory and known to be impaired in the early stages of AD [140]. IAPP, a protein related to diabetes type 2 pathology, was also found to form aggregates in the presence of zinc in pancreatic β-cells, while iron deposits were characterized in the Lewy bodies formed in PD [141,142].

Metal ions from the d-block generally show an inducer effect on protein/peptide fibrillization pathways in vitro. However, discrepancies exist between the final aggregation states that can be reached. The nature of the metal ion, of the amyloid system and of the peptide/metal stoichiometry are critical parameters, as illustrated in the following examples. Cu(II) was reported to significantly reduce the lag phase and promote fibril formation for α-synuclein [143]. However, Cu(II) was recently shown to inhibit IAPP fibrillization and to promote an alternative aggregation pathway towards the formation of toxic globular oligomers [144]. For the Aβ peptides, sub-equimolar concentrations of Cu(II) were found to promote the formation of amyloid nuclei, and hence to increase the rate of fibrillization by decreasing the lag phase [145,146]. However, at super-stoichiometric and equimolar concentrations of Cu(II) and Zn(II), the fibrillization process was inhibited and Aβ then formed amorphous aggregates, although slow conversion to amyloid fibrils was still observed for the latter condition [147–149]. Aβ annular protofibrils were observed in the presence of excess concentrations of iron, while manganese did not show a significant binding affinity to Aβ peptides [150,151]. Similarly to GAGs, the effect of metal ions on fibrillization processes has to be considered on a case-by-case basis. Depending on the amyloid system, the type of ion and the stoichiometry, we see that induction, inhibition or promotion of alternative pathways of the fibrillization process can occur.

Metal ions from the d-block can interact with numerous binding sites on proteins via coordination or electrostatic interactions, through affinities for side chains, backbone carbonyl groups or terminal amine groups. For instance, Aβ major binding sites include the side chains of histidine (His6, 13, 14), tyrosine (Tyr10), aspartate (Asp1, 7 and 23), glutamate (Glu3, 11 and 22) and methionine (Met35), together with carbonyl groups from the peptide backbone [52,152,153]. From a mechanistic point of view, the dynamics of metal ion binding to amyloidogenic species are reported to be faster than the conformational or oligomeric conversions of the fibrillization process. Structural and kinetic characterizations supported Zn(II) to promote the Aβ11–28 dimerization by bridging peptides, and hence acting as a catalyst of the first oligomerization steps towards fibrillization. However, transient Zn(II) binding and metal ion exchange between early aggregates and soluble peptides was proposed, as metal–peptide fibrils induced peptide fibrillization but not metal–peptide fibril propagation [154]. Transient metal binding to amyloidogenic peptide/protein species and fast exchange rates of the metal ions between intra- and intermolecular peptide-binding sites are currently thought to dominate the metal–amyloid interactions [155] (figure 7).

Figure 7.

Mechanisms of metal-induced fibrillization, without transient binding (a) or with transient binding (b). (Reproduced with permission from Faller et al. [155] (Copyright © 2016 American Chemical Society).)

The strong binding affinity of typically noble metals for the thiol moiety of cysteine has been widely exploited in hybrid and inorganic nanomaterial synthesis from amyloid systems that bear this residue. Numerous and diverse nanotechnology applications based on cysteine–metal interactions in amyloid systems have emerged in recent years. Hybrid biomaterials such as ultra-light gold–amyloid hydrogels—known as aerogels [156]—conductive gold–amyloid films [157] or amyloid–metal hybrids for continuous flow catalysis [158] have been successfully fabricated using this chemical affinity. Amyloid nanofibrils with cysteine residues have also been shown to enhance the transfection yield of gold nanoparticles [159], or to be useful templates for directing the synthesis of metallic nanofibres, nanorods and nanoparticles, as extensively reviewed elsewhere [160–164]. Recently, Bolisetty & Mezzenga [165] used the strong affinity between heavy metals and amyloids presenting cysteine residues on their surface to create a unique water purification system. Specifically, hybrid amyloid–carbon membranes were created to selectively trap heavy metal ions and other contaminants from water samples with remarkable efficiency. Considering the variety of hybrid or inorganic nanomaterials that have been successfully synthesized to date using cysteine-containing amyloid systems, it seems very likely that commercial versions of such materials should be available in the not too distant future.

6. Conclusion: towards amyloid disease therapy and bionanomaterial design

Although the molecular mechanisms involved in amyloid cytotoxicity are still unclear and no curative treatments are available, strategies aiming at amyloid disease therapy are being intensively explored. These mainly focus on interfering with the fibrillization process, via trapping non-toxic intermediates or stabilizing the protein native conformation using xenobiotics. Small xenobiotic molecules have been shown to directly interact with the fibrillization pathway, specifically the growth of the extended β-sheet networks, including a wide range of phenolic compounds, such as curcumin or polyphenols [166,167]. The corresponding molecular interactions still need to be unravelled, and the potential for therapy confirmed, as such compounds may also stabilize toxic oligomers, given the complexity of the amyloid fibrillization pathway. The strategy consisting of stabilizing the protein native conformation is promising, particularly for transthyretin amyloidosis. This condition can lead to heart failure due to transthyretin fibrillogenesis in this organ. Compounds preventing the dissociation of the native tetramer are currently being considered in clinical trials [168,169]. Therapeutic proteins may further emerge from the recently evidenced regulation of the functional bacterial curli fibrillization within their host by protein chaperones [26]. These delay amyloid nuclei formation until fibrillization is needed for bacterial biofilm formation, by direct electrostatic interactions with the protein monomers. Interestingly, such chaperones were shown to retain a similar inhibitory activity in vitro on α-synuclein fibrillization, potentially boding well for PD treatments [170]. However, in order for such treatments to be effective, there is a requirement for new diagnostic tools allowing early (preferably pre-symptomatic) disease detection, as the chaperones cannot reverse amyloidogenesis. Advances in the understanding of the amyloid cytotoxicity mechanisms would therefore support the emergence of new therapeutic strategies.

Their biomimetic morphology and material properties make networks of amyloid fibrils assembled from synthetic peptides attractive materials for the fabrication of scaffolds for tissue engineering and other biomaterial applications. The majority of systems studied to date exist as passive structural scaffolds with limited applications. Clinically beneficial scaffolds should be able to direct cell responses and not just support cell growth [171]. To enable this, the fibrils should display specific cell recognition peptide sequences, growth factors and cytokines on their surface. Furthermore, to allow for well-defined stoichiometric decoration of the fibril surface, it is beneficial to include these moieties into the assembling peptides as opposed to functionalizing the scaffolds post-assembly. While there have been a few examples of synthetic amyloidogenic peptides with additional cell recognition sequences [102,103], the synthesis of such amyloid-forming systems is far from routine. An increased understanding of amyloid assembly pathways and how they will be affected by the introduction of additional biomolecules will greatly facilitate the rational synthesis of scaffolds that can control cellular behaviour. For example, implantable amyloid scaffolds could be seeded with a patient's own stem cells, and growth factors on the surface of the scaffold can then direct cell differentiation into specific lineages to promote wound healing or tissue regeneration [105].

As a final comment, although outside the scope of this review, amyloid nanofibril functionalization with diverse chemical groups or biomolecules is to be mentioned for its high potential to provide novel functional bionanomaterials, as reviewed elsewhere [163,172]. For instance, semi-conductive peptide nanofibrils [173] or enzyme immobilization on protein nanofibrils have been successfully obtained via specific chemical functionalization steps [174]. Diverse applications may emerge from this research, from novel nanomaterials for nanoelectronics to biosensing in medical sciences.

Authors' contributions

N.P.R., C.E.C. and C.V. initiated the project. D.D. performed the initial literature search. D.D., N.P.R., C.E.C. and C.V. wrote the paper. N.P.R. created the original diagrams.

Competing interests

The authors declare no competing interests.

Funding

D.D. is the recipient of an RMIT PhD scholarship. C.E.C. is the recipient of an ARC DECRA fellowship DE160101281. N.P.R. is funded through the ARC Training Centre for Biodevices at Swinburne University of Technology (IC140100023).