Amyloid formation involves three pathological protein transformations: structural, from natively folded to the cross-β conformation; biophysical, from soluble to insoluble; and biological, from functional to non-functional [68]. The cross-β conformation buries the once-functional domains of the protein within the steric zipper architecture, which makes amyloids extremely stable [14], and, consequently, relatively inert. Additionally, the uncontrolled phase transition leads to loss of protein solubility and colloidal stability resulting in precipitation into plaques, which further buries any potential unpaired side chains via hierarchal self-interaction (see above). Within plaques, amyloid protofilaments and fibrils adopt different polymorphic morphologies depending on environmental conditions. However, since all polymorphs are based on the same cross-β conformation, where the functional side chain domains are sequestered, they are expected to have similar, generic amyloid properties in terms of stability, insolubility, and low reactivity. Furthermore, the loss of protein solubility and colloidal stability favors precipitation and cluster formation over propagation of single fibrils, which requires colloidal dispersion. This is supported by clinical findings that plaques (heterogenous fibrillar clusters) are the hallmarks of amyloid pathologies.

With the high stability and low reactivity, amyloids pose little direct toxicity unless they physically remodel a tissue, for example, in the rare cases of systemic amyloidosis such as immunoglobulin light chain amyloidosis and transthyretin amyloidosis [74,75]. This is especially true for amyloid accumulation within the muscle tissue (e.g., cardiac amyloidosis), where amyloid infiltration physically impair muscle contractility [76]. However, in many other cases, amyloids exist as a benign mass similar to other benign masses, such as fibromas and lipomas. In insulin-derived amyloidosis, for example, repeated injection of insulin subcutaneously in the same spot leads to the creation of insulin amyloid lumps in some diabetic patients [77]. Despite the benign nature of such lumps, patients lose the ability to control glucose levels due to insulin sequestration in the form of amyloid aggregates [78]. Patients are, therefore, instructed to change the location of insulin injections to avoid local aggregation. In this case, the toxicity due to amyloid aggregation is due to loss-of-function (LOF) of the injected insulin and not due direct toxicity from the amyloid mass. Pathogenesis due to LOF is also demonstrated in the case of p53 amyloid formation. P53 is a tumor suppressor protein whose dysregulation or inactivation is involved in more than 50% of all cancers [79]. It has been shown that p53 can form amyloid fibrils leading to enhanced cell proliferation due to its sequestration and LOF [80,81]. These findings clearly indicate that amyloids are not necessarily cytotoxic as they can enhance, not impair, cell proliferation via a LOF mechanism of p53. LOF is also the mechanism behind many phenotypes in yeast due to amyloid formation [82]. For example, amyloid formation of Sup35, which is an essential translation termination factor, induces lethality due to LOF as a result of its sequestration in the amyloid state [83]. Such an outcome can be reversed by supplying the yeast with a modified version of Sup35, where the residues more prone to amyloid formation are removed, while the domains involved in translation termination are maintained [84]. This replacement approach to overcome amyloid LOF toxicity is also utilized clinically in the treatment of diabetes mellites by using pramlintide, which is a less aggregating analogue of the peptide hormone amylin, whose amyloid aggregation in the pancreas and depletion in the circulation is common among diabetic patients [85]. The relatively benign nature of amyloid plaques can also be seen in neurodegenerative diseases such as AD, where up 30% of individuals who have plaques in their brains are cognitively normal [86]. We have recently shown that higher levels of soluble Aβ42 are associated with normal cognition and preservation of brain volume among amyloid positive individuals, regardless of and despite increasing levels of brain amyloid, indicating that LOF of the soluble Aβ42 is more detrimental to neurons than direct gain-of- function (GOF) toxicity from plaques [87]. This suggests that a replacement approach might also be feasible for AD treatment and other neurodegenerative diseases, an alternative to continuing with anti-amyloid strategies, which have invariably failed [88].

“Toxic oligomers”? Oligomers have been postulated to explain the lack of association between amyloid plaque load and toxicity, especially in AD. The term oligomers denotes low and medium molecular weight aggregates that are assumed to mediate the amyloid toxicity [89]. However, the evidence of their toxicity has been shown in vitro, not clinically, and the clinically relevant toxic oligomer remains unknown [90]. Moreover, under supersaturated conditions, which are necessary for amyloid formation, the distinction between oligomers and nuclei is hard to make, since the formation of any cluster stable enough will trigger phase transition into fibrils, whereas unstable clusters will dissociate back into monomers (see above). This has been demonstrated experimentally, where the majority of oligomers were shown to dissociate into monomers, in good accordance with the classical nucleation theory [91]. Moreover, the reduction of soluble Aβ42 levels during the disease course reduces the substrate for the oligomers, questioning the long-term effect of such species. This is supported by the fact that attempts to quantify oligomeric species of Aβ found that they are less in AD patients compared to controls [92,93,94].

“Ratios”. Despite the progressive decrease in the absolute levels of soluble Aβ42 in AD, the ratio of Aβ42 relative to other shorter versions of the peptide, such as Aβ40, was hypothesized to increase the likelihood of Aβ42 forming aggregates due to its more amyloidogenic nature [95]. However, amyloid aggregation is dependent on supersaturation, which is dependent on the absolute concentration of the peptide. Decreasing peptide concentration will decrease, not increase, its propensity to form any type of aggregates irrespective of its relative levels compared to other peptides. Moreover, it has been recently demonstrated that CSF Aβ42/Aβ40 ratio also decreases during the course of AD [72].

Another indication on the importance of LOF as a pathogenic mechanism in neurodegenerative diseases is the fact that animal models where the amyloidogenic proteins were knocked out or down display phenotypes that resemble those obtained by protein overexpression and aggregation. This has been demonstrated for Aβ42 [96,97] in AD, α-synuclein in Parkinson’s disease [98,99,100], and other neurodegeneration-related proteins that form amyloids such as Tau [101], PrP [102], SOD1 [103], and TDP43 [104]. The similarity of phenotype in both the presence and absence of aggregates can only be explained by LOF mechanism, where the sequestration of protein due to aggregation mimics the effects of gene knock down. Additionally, in many cases, the phenotype can be rescued by restoration of normal soluble levels of these proteins [97,99]. For further discussion on LOF toxicity, we refer the reader to our earlier reviews [68,105].