Prions, amyloids, and RNA: Pieces of a puzzle

ABSTRACT

Amyloids are protein aggregates consisting of fibrils rich in β-sheets. Growth of amyloid fibrils occurs by the addition of protein molecules to the tip of an aggregate with a concurrent change of a conformation. Thus, amyloids are self-propagating protein conformations. In certain cases these conformations are transmissible / infectious; they are known as prions. Initially, amyloids were discovered as pathological extracellular deposits occurring in different tissues and organs. To date, amyloids and prions have been associated with over 30 incurable diseases in humans and animals. However, a number of recent studies demonstrate that amyloids are also functionally involved in a variety of biological processes, from biofilm formation by bacteria, to long-term memory in animals. Interestingly, amyloid-forming proteins are highly overrepresented among cellular factors engaged in all stages of mRNA life cycle: from transcription and translation, to storage and degradation. Here we review rapidly accumulating data on functional and pathogenic amyloids associated with mRNA processing, and discuss possible significance of prion and amyloid networks in the modulation of key cellular functions.

INTRODUCTION

Pathological extracellular deposits, which were most likely amyloidoses, were described in liver, kidney and other human organs as early as in the 17th century.^1,2 The term amyloid was coined in 1854, after the first chemical tests of extracellular waxy deposits in the brain and spinal cord: Rudolf Virchow detected their staining with iodine and postulated that they were composed of cellulose or starch.³ In just five years the prevailingly polysaccharide nature of such deposits was refuted. Based on their high nitrogen content, Friedreich and Kekule concluded that the deposits were primarily proteinaceous,⁴ but chose not to argue with the name. The iodine staining was eventually explained by the presence of polysaccharides in the form of glycosaminoglycans in most amyloids.⁵ Present-day evidence upholds that proteins determine the formation of amyloid deposits and remain their key components, and current chemical tests for amyloids are based on dyes that bind to amyloid-specific protein structures. These tests include apple-green birefringence in polarized light following staining with the Congo Red dye,⁶ or a shift of fluorescence spectrum upon binding of a fluorescent dye Thioflavine-T.^7-9 These notable effects occur due to the presence of highly ordered amyloid fibrils where protein molecules form β-strands running perpendicular to the lateral axis of the fibril.^10,11 Such architecture of amyloid fibrils also results in their characteristic “cross-β” pattern in X-ray diffraction,¹⁰ as well as an unusually high resistance to treatment with ionic detergents and,^12-14 in some cases, with proteinases.¹⁵ The detailed structure of the fibers, and amino acids involved in the formation of β-strands, are known only for some amyloids, which were prepared in vitro from ectopically expressed and purified proteins. However, considering the frequent occurrence and fundamental similarity of amyloids, the term “amyloid” is also used, and the abovementioned structure is presumed, for a broad range of extracellular deposits in tissues, intracellular inclusions and protein aggregates satisfying at least one criterion of amyloids, as long as there is no evidence for an alternative structure.

To date, at least 32 proteins are known to form pathological amyloids, which have been associated with dozens of incurable diseases in humans and animals.^16,17 However, the other side of the coin was revealed in 2000, when protective envelope of the egg of silk moth Bombyx mori¹⁸ and hydrophobins of basidiomycete Schizophyllum commune¹⁹ were proven to have amyloid properties. These studies demonstrated that amyloids can be not only pathogenic, but also functional, and were soon followed by a number of reports describing the biological importance of other amyloids. Eventually, functional amyloids were discovered in all domains of life. They participate in the formation of biofilms in Archaea,²⁰ as well as in different species of Bacteria, for which at least six groups of functional amyloids have been identified.^17,21 In Eukarya, amyloids possess a broad spectrum of functions including, in addition to the aforementioned, deposition of hormones,²² regulation of melanin polymerization,²³ modulation of long-term memory,²⁴ anti-viral response,²⁵ and programmed necrosis.²⁶

Unexpectedly, the path of amyloid research intersected with studies of infectious protein conformations, or prions. Extreme resistance of PrP^Sc to protease digestion was reported right upon the identification of this protein as the only component consistently co-purifying with the infectivity of transmissible spongiform encephalopathies.^15,27 Furthermore, histopathological studies of brains of scrapie-infected animals and Creutzfeldt-Jakob disease patients revealed the accumulation of PrP^Sc in congophilic amyloid plaques.^28,29 Finally, in vitro, PrP formed rod-like aggregates that exhibited apple-green birefringence upon binding Congo Red.³⁰ With these characteristics being major hallmarks of amyloids, experimental evidence indicated that PrP was an amyloidogenic protein. Strikingly, while the infectious PrP^Sc was clearly associated with the aggregated amyloid state of the PrP protein, the non-prion form of PrP, PrP^C, was soluble and non-amyloid. The explanation how amyloid structure could be so efficiently utilized by an infectious protein conformation was provided by the nucleation-polymerization model. This model postulates that prions appear through a nucleation event involving a conformational conversion and oligomerization, grow through the addition of protein monomers to the nucleus, with concurrent conformational conversion, and infect through the transmission of fragments of prion aggregates called seeds.^31,32

Thus, once other prions were later discovered in Saccharomyces cerevisiae, Podospora anserina and, eventually, in higher Eukaryotes, the amyloid structure of prion conformations did not come as a surprise. The highest number of prions has been identified in fungi, and particularly in baker's yeast S. cerevisiae. While opinions still differ on whether the majority of yeast prions are functional epigenetic modifiers, egoistic elements or diseases, most of them do not have a significant negative effect on fitness and can be maintained in yeast populations for a long time, or even indefinitely.^33,34 To date there are nine known amyloid-based yeast prions, [PSI⁺]³⁵ (for which structural protein is Sup35),³⁶ [URE3]³⁷ (Ure2),³⁶ [PIN⁺] (Rnq1),^38,39 [SWI⁺] (Swi1),⁴⁰ [OCT⁺] (Cyc8),⁴¹ [MOT3⁺] (Mot3),¹⁴ [ISP⁺] (Sfp1),⁴² [NUP100⁺] (Nup100),⁴³ [MOD⁺] (Mod5),⁴⁴ as well as two functional prion-like amyloids: [PUB1] (Pub1) and [PUB1 / SUP35] (Pub1 and Sup35).⁴⁵ Also, the [HET-s] prion is an amyloid controlling heterokaryon incompatibility in filamentous fungus Podospora anserina.^46,47 The only known prions that do not possess amyloid properties are [β]⁴⁸ and [GAR⁺]^49,50 from S. cerevisiae. For [β] the principle of self-propagation is completely different: this is a protease that propagates through covalent autoactivation. As for [GAR⁺], while the Pma1 and Std1 proteins have been associated with the formation of this prion, the involvement of other proteins and an amyloid structure cannot be excluded. Thus, it is currently unclear to what extent prions and amyloids should be considered as two overlapping sets, or if all prions not functioning through covalent automodification are amyloids.

The β-sheet-based structure of amyloid fibrils engenders great diversity both in pathogenic and functional aspects. It should be acknowledged that it is still not completely clear what makes a protein amyloidogenic, but some conclusions can already be drawn from studies of amyloidogenic sequences. The amyloid-forming proteins identified to date possess regions with “unusual” amino acid composition. Such regions, depending on the method used for their prediction, are called “compositionally biased regions” (CBRs)⁵¹ or “low complexity regions” (LCRs).⁵² In essence, the term CBR implies that such region is rich or poor in particular residues compared to the average occurrence frequency of these residues in the proteome.⁵¹ So far two types of CBRs have been described for amyloid-forming proteins: (i) sequences rich in glutamine (Q) and/or asparagine (N);^51,53 and (ii) sequences rich in hydrophobic and non-polar residues such as I, W, F, Y, L and V.^54,55 The term LCR implies that a region contains little diversity in its amino acid composition.⁵² To a certain extent, low complexity is an expected consequence of significant enrichment with one or several amino acids that reduces the representation of other amino acids. The following classification of LCRs can be suggested in connection with amyloid formation: (i) LCRs lacking, or exhibiting only loose clustering of overrepresented amino acids; (ii) LCRs with extended interrupted or uninterrupted runs of homo-amino acid repeat tracts; (iii) LCRs with periodic sequences, such as tandem oligopeptide repeats. CBRs / LCRs affect amyloid formation in two different ways: (i) upon initial protein folding they maintain the region of the protein as intrinsically disordered and thus available for a conformational switch, and (ii) they promote amyloid formation through the formation of intermolecular β-strands. The first is best achieved by CBRs / LCRs rich in polar and charged residues, which increase the solubility of the protein. Amino acid composition of such regions appears to be more critical than the exact amino acid sequence.⁵⁶ The second is facilitated by monotonous and repetitious sequences, including those rich in hydrophobic residues,⁵⁷ and in this case the position of each residue is very important: the same residue may either promote, or block the formation of amyloid, depending on the context.⁵⁸ Notably, analysis of amyloidogenic proteins reveals that some of them carry both types of CBRs / LCRs.⁵⁹

Computational prediction of amyloid properties is efficient for short peptides. The false positive rate for full-length proteins is much higher,⁶⁰ although some recently developed algorithms provide up to 70% of true positive predictions for full-length proteins.⁶¹ The difference in the efficiency of predictions is because in real proteins amyloidogenic regions are interspersed by non-amyloidogenic ones in primary sequences. Currently, there are no reliable algorithms that analyze the interplay of amyloidogenic sequences with other regions of a protein and determine the ability or inability of an amyloidogenic region to drive aggregation under physiological conditions.

Noteworthy, amyloidogenic regions can exhibit complex organization, where one protein may contain not one, but numerous amyloidogenic determinants. For example, Rnq1, a structural protein for the [PIN⁺] prion,³⁸ has four Q/N-rich regions. Each of these regions alone can promote the aggregation of Rnq1 in vitro, and one common Q/N region is sufficient for the transmission of the prion state between Rnq1 fragments, though the overall conformation of the [PIN⁺] prion is determined by the cooperative action of all four determinants.⁶² It has also been suggested that Rnq1 encompasses non-Q/N-rich amyloidogenic regions, both interspersing the Q/N-rich determinants and located outside of the prion domain, and that these regions differentially affect the maintenance of [PIN⁺] variants.⁶³ Analogously, the prion domain of the CPEB3 protein carries two aggregation domains with non-identical roles in amyloid formation, which are separated by a module regulating aggregation by affecting CPEB3 interaction with the actin cytoskeleton.⁵⁹

One intriguing feature of the amyloid-forming proteins identified to date is a significant overrepresentation of mRNA-processing proteins. This peculiarity was first noted for Q/N-rich proteins of the yeast Saccharomyces cerevisiae,⁵³ and was later upheld for confirmed prions. Indeed, nine prion- and amyloid-forming proteins of S. cerevisiae and at least 13 amyloid-forming proteins of higher eukaryotes are known to possess mRNA-processing functions. In some cases, the amyloid state of these proteins is neutral or functional, in others – pathogenic. In this review we analyze biological diversity of amyloids formed by proteins involved in mRNA processing, summarize their functional and pathological roles, and discuss a possible significance of prion and amyloid networks for the modulation of key cellular processes. Data on biological diversity of such amyloids, their functions, and processes, in which they are involved, are summarized in Table 1 and Figure 1. Although many amyloid-forming proteins affect mRNA processing in many ways, for the purpose of this review we tried to group them according to the stages of the mRNA life cycle in which they are mostly involved: (i) transcription, (ii) mRNA turnover, and (iii) translation.

TABLE 1.

RNA-modulating prions, amyloids, and amyloidogenic proteins

Gene / Protein	Amyloid (Prion)*	Functions of the protein**	Effects of prion or amyloid state	Organism	Compositionally biased regions***
URE2 / Ure2	[URE3]	Transcriptional regulator of the nitrogen catabolism	Blocks interaction of Ure2 with Gln380 that causes preference of poor nitrogen sources	S. cerevisiae	N, Q/N
SWI1 / Swi1	[SWI+]	Transcriptional regulator, subunit of SWI/SNF chromatin remodeling complex	Inhibits growth on the media with non-fermentable carbon sources,40 provides resistance to benomyl14	S. cerevisiae	A, N, Q, Q/N, T
SUP35 / Sup35	[PSI+]	Translation release factor eRF3, mRNA decay	Reduces efficiency of translation termination178	S. cerevisiae	E, G, K, Q, Q/N, Y
MOT3 / Mot3	[MOT3+]	Transcription factor modulating mating, carbon metabolism, and stress response	Regulates flocculation; induction by ethanol and elimination by hypoxia85	S. cerevisiae	A, H, N, P, Q, Q/N
CYC8 / Cyc8	[OCT+]	Transcriptional repressor and activator, subunit of Cyc8-(Ssn6)-Tup1 complex	Inhibits growth of the strains bearing cyc1 deletion on the media with non-fermentable carbon source lactate41	S. cerevisiae	A, E, P, Q/N, Q
SFP1 / Sfp1	[ISP+]	Transcription regulator of genes encoding ribosomal proteins; regulates response to nutrients and stress	Antisuppression of nonsense mutations;87 enhances resistance to cycloheximide and paromomycin42	S. cerevisiae	A, D, H, N, Q, Q/N, T
PUB1 / Pub1	[PUB1]	Stress granule assembly, 3′UTR mRNA-binding	Unclear; aggregates co-localize with P-bodies / stress granules45	S. cerevisiae	M, N, Q, Q/N
PUB1 and SUP35 / Pub1 and Sup35	[PUB1 / SUP35]	Stress granule assembly / Translation release factor eRF3	Formation of microtubule-associated cytoskeleton-associated complex carrying translational machinery and important for maintaining the integrity of the microtubular cytoskelton45	S. cerevisiae	E, G, K, M, N, Q, Q/N, Y
TIA1 / TIA1	[TIA1]	Stress granule assembly, 3′UTR mRNA-binding	Amyloid-like fibrils involved in stress granule assembly111	M. musculus	Q
NUP100 / Nup100	[NUP100+]	Subunit of nuclear pore complex; nucleocytoplasmic transport	Unknown, phenotypic manifestation is absent43	S. cerevisiae	G, N, Q, Q/N, S, T
TP53 / p53	p53, mutant	Transcriptional repressor	Amyloid-like oligomers and fibrils associated with carcinogenesis104,105	H. sapiens	A, P
HTT / huntingtin	huntingtin, mutant	Transcriptional regulator	Pathological amyloid-like inclusions associated with Huntington's disease91	H. sapiens	D, L, P, Q, Q/N, V
ATN / atrophin-1	atrophin-1, mutant	Transcriptional repressor	Pathological amyloid-like inclusions associated with dentatorubral pallidoluysian atrophy94	H. sapiens	D, E, G, H, P, Q, Q/N, R, S
TNRC6A / TNRC6A	TNRC6A	RNA-binding, component of P-bodies, involved in the regulation of post-transcriptional gene silencing through the RNA interference	Fibrils in P-bodies, function is unclear120	H. sapiens	G, K, N, P, Q, Q/N, S, T, W,
DDX6 / DDX6	DDX6	RNA helicase found in P-bodies and stress granules, involved in translation suppression and mRNA degradation	Fibrils in P-bodies, function is unclear121	H. sapiens	Q
TDP43 / TDP43	TDP43, mutant	Transcriptional repressor, component of stress-granules	Pathological inclusions associated with frontotemporal dementia and amyotrophic lateral sclerosis127-129	H. sapiens	G, N
FUS / FUS	FUS, mutant	RNA-binding, component of hnRNP-complex	Pathological inclusions associated with amyotrophic lateral sclerosis and polyglutamine diseases131,132	H. sapiens	G, Q, R, S, Y
HNRNPA1 / hnRNPA1	hnRNPA1, mutant	RNA-binding, involved in the packaging of pre-mRNA into hnRNP particles, transport of poly-A mRNA from the nucleus to the cytoplasm	Pathological, multisystem proteinopathy-associated intracellular amyloid-like inclusions136	H. sapiens	G
HNRNPA2B1 / hnRNPA2B1	hnRNPA2B1, mutant	RNA-binding, associated with pre-mRNAs in the nucleus	Pathological, multisystem proteinopathy-associated intracellular amyloid-like inclusions136	H. sapiens	G, Y
CPEB / CPEB	CPEB	RNA-binding, binds U-rich sequence located in the 3′ untranslated regions of mRNAs, thus either promoting, or inhibiting translation	Functional amyloid involved in the control of long-term memory; induction by 5HT24,159	A. californica	P, Q, Q/N, S
orb2 / Orb2A	Orb2A	see above (CPEB)	see above (CPEB); induction by dopamine, octopamine, or tyramine161,169	D. melanogaster	G, H, M, Q, Q/N, S
CPEB3 / CPEB3	CPEB3	see above (CPEB)	see above (CPEB); induction by dopamine, glutamate, or glycine59,162,163	M. musculus	A, P, Q, Q/N

^*Column indicates the special designation of amyloid (or prion, in the case of yeast prions).

^**Based on data from the NCBI Gene database (http://www.ncbi.nlm.nih.gov/gene/), Saccharomyces Genome Database (http://www.yeastgenome.org/), and cited studies.

^***Overrepresented amino acids are shown. CBRs were predicted by SARP algorithm,¹⁷⁹ and the probability thresholds were set to 10⁻⁶ for single residue CBRs and 10⁻¹² for multiple residues CBRs to reduce false positives. CBRs for multiple residue were predicted for the Q/N pair only.

FIGURE 1.

RNA-modulating prions and amyloids, and cellular processes in which they are implicated. Shown are cellular processes or protein complexes, which are associated with corresponding prions and amyloids. Arrows connect consequent stages of the mRNA life cycle.

I. Prions, Amyloids, and The Modulation of Transcription

Considerable data on the modulation of transcription by amyloid-forming proteins has been obtained for prions of the yeast S. cerevisiae. Currently, the list of prions for which prion-forming proteins possess an activity of transcriptional regulators includes [URE3], [SWI⁺], [MOT3⁺], [OCT⁺], and [ISP⁺]. In addition, the [NSI⁺] prion-like determinant,^64,65 for which the prion-forming protein is unknown, was shown to slightly modulate the levels of mRNAs of several genes.^66,67 Prion-forming proteins of the corresponding prions serve different roles in transcriptional regulation. Ure2 is a component of a system regulating nitrogen catabolism through the modulation of the localization of the GATA transcriptional activator Gln3.⁶⁸ Swi1 is a subunit of SWI/SNF chromatin remodeling complex, which is required for the transcription of many genes including those controlling metabolism of sugars and mating type switching.⁶⁹ Mot3 is a transcriptional repressor and activator involved in the regulation of mating, carbon metabolism, stress response, and controls a complex cell wall remodeling program during the adaptation to anaerobiosis.⁷⁰ Cyc8, the prion-forming protein of [OCT⁺], is a component of Cyc8/Tup1 complex that represses the transcription of over 150 genes.^71,72 Interestingly, while Mot3 contributes to the recruitment of the Cyc8/Tup1 complex to repress transcription, Cyc8/Tup1 is involved in the recruitment of the SWI/SNF complex to activate transcription.⁷³ Finally, Sfp1 is a transcriptional activator of the genes coding for ribosomal proteins.⁷⁴

All prions within the aforementioned group form aggregates with amyloid properties. However, an extensive characterization of amyloid fibrils including the Congo Red assay, electron microscopy and solid state NMR structural analysis of in vitro-made amyloid has been completed for Ure2 only.^75-77 Swi1 and Mot3 were shown to form SDS-resistant aggregates in the yeast cytosol,¹⁴ and amyloid fibers assembled in vitro from the ectopically expressed protein fragments corresponding to prion-forming domains of these proteins caused a shift in fluorescence spectrum when stained with Thioflavin-T.^14,78 For Cyc8 fused to YFP, fluorescent microscopy revealed formation of aggregates in [OCT⁺] cells.⁴¹ Currently, there is no evidence to suggest that they are detergent-resistant,⁴¹ although the Q/N-rich region of Cyc8 forms SDS-resistant aggregates.¹⁴ Sfp1 forms aggregates in the nuclei of [ISP⁺] cells. Such localization is unique among known yeast prions.⁴² Intriguingly, the Q/N rich region of Gln3, a protein interacting with Ure2 and playing the central role in the [URE3] manifestation, forms amyloid aggregates,¹⁴ although this propensity was not tested for the full-length protein.

Prions were originally described as deleterious factors.⁷⁹ Accordingly, when prions were discovered in yeast, it was postulated that proteins in the prion conformation are inactivated, and the phenotype of prion state corresponds to a deletion or an inactivating mutation in the gene encoding the prion-forming protein.³⁶ Initial experimental evidence for [URE3] was consistent with this notion. However, further in-depth studies of [URE3] and recent studies of [SWI⁺], [MOT3⁺], and [ISP⁺] indicate that the situation is not that simple. Formation of [URE3] is known to prevent the interaction of Ure2 with Gln3,⁸⁰ which, through a cascade of transcriptional factors,^81,82 makes yeast to catabolize poor nitrogen sources, such as ureidosuccinate, in the presence of rich nitrogen sources. The original phenotypic assay for [URE3] was based on this disruption of Gln3 regulation; and thus, the formation of [URE3] could be interpreted as inactivation of Ure2.³⁷ However, in addition to transcriptional regulation, Ure2 possesses glutathione S-transferase activity,⁸³ and this function persists in both [URE3] and [ure-0] states.⁸⁴ Analogously, the appearance of [SWI⁺] leads to the inability to grow on non-fermentable carbon sources, which is characteristic of SWI/SNF inactivation. However, unlike the swi / snf mutants, yeast bearing the [SWI⁺] prion are not sensitive to 0.3 M LiCl or 1 M NaCl.⁴⁰ [MOT3⁺] causes several phenotypes, including flocculation, that are not observed upon the deletion of the corresponding gene.⁸⁵ Most strikingly, Sfp1 preserves its function of the transcriptional regulator in [ISP⁺] strains.⁸⁶ Furthermore, while the SFP1 deletion leads to slow growth and smaller cell size, prionization of Sfp1 does not.⁴² [ISP⁺] even has the opposite effect on fitness: it increases growth rate relative to the [isp⁻] strain.⁸⁷ So, prionization of transcriptional regulators can not only inhibit, but also modulate their functional activities.

Some of the prions discussed in this section also provide a beneficial resistance to various drugs and compounds: [SWI⁺] - to benomyl, [MOT3⁺] - to calcofluor white and Congo Red,¹⁴ and [ISP⁺] - to cycloheximide and paromomycin, but it is unclear whether these benefits are ever used by wild yeast.⁴²

So far, of all these prions, the presence in natural yeast isolates has only been reported for [MOT3⁺]. ⁴³ Notably, a specific mechanism of induction and elimination driven by changes in environmental conditions has been described for this prion: increasing ethanol concentration promotes [MOT3⁺] formation, while hypoxia eliminates it by repressing the expression of the MOT3 gene.⁸⁵ Such changes in environmental conditions do occur during fermentation, and, taking into consideration that the appearance of [MOT3⁺] causes FLO11-mediated flocculation, which is important for survival, this prion is apparently beneficial, although, in fact, this has not been confirmed outside of the laboratory.⁸⁵

[URE3] does not provide any apparent advantages to yeast, and some strains or conformational variants of this prion significantly inhibit vegetative growth.⁸⁸ Also, [SWI⁺] manifests not only beneficial, but also harmful phenotypes: it inhibits vegetative growth on media containing non-fermentable carbon sources.⁴⁰ This might mean that these prions are egoistic elements or even that they are harmful. However, we cannot exclude the possibility that these prions provide a survival advantage under certain conditions that occur rarely and which we do not yet know, and thus act as bet hedging prions.³³

In mammals, mutant huntingtin and p53 are examples of potentially harmful amyloids formed by proteins involved in transcription. Mutant huntingtin protein containing an expanded poly-Q tract in its first exon⁸⁹ forms intranuclear and cytoplasmic inclusions⁹⁰ that have some amyloid characteristics⁹¹ and causes a lethal neurodegenerative disorder known as Huntington's disease.^92,93 A hallmark of Huntington's disease and a set of disorders caused by the expansion of poly-Q tracts in several other proteins, such as ATN (atrophin-1), is a significant change in the transcription of different genes that is presumably associated with pathogenesis.⁹⁴ Currently, the molecular mechanism, through which these changes occur, is unclear, but several interesting hypotheses are worth mentioning. One is based on the finding that inclusions of huntingtin and atrophin-1 sequester some poly-Q-containing transcription factors:^95-97 it was suggested that such sequestration might interfere with the functional activity of the corresponding transcription factors.⁹⁴ This model is still under consideration, although incomplete recruitment of the transcription factors by the inclusions⁹⁸ is viewed as an argument against the sequestration hypothesis. Other explanations do not implicate inclusions formed by proteins with poly-Q expansions. For example, mutant and wild-type huntingtin have been shown to have direct DNA-binding sites, and these sites are different.⁹⁹ Therefore, mutant huntingtin could perturb normal transcription by preventing or modulating the binding of normal transcription factors to DNA. Alternatively, there is evidence that transcriptional deregulation could be due to the presence of mRNAs carrying the expansions of CAG codons coding for the poly-Q stretches.¹⁰⁰

The tumor suppressor p53, in its normal form, is a tetrameric transcription factor that blocks proliferation by inducing cell-cycle arrest.¹⁰¹ Mutations in TP53, the p53-encoding gene, often result in tumor progression,¹⁰² which occurs either due to the functional inactivation of p53, or the sequestration of the wild type protein by its dominant-negative mutant form.¹⁰³ In the latter case, such a dominant-negative form can be manifested by amyloids of mutant p53.¹⁰⁴ The exact role of p53-containing amyloids in carcinogenesis is currently unclear, though these amyloids are detected in samples of breast cancer¹⁰⁵ and co-localize with wild-type p53 in several cell lines.¹⁰⁴ In summary, there is evidence that proteins involved in transcriptional regulation form amyloids in mammals. Amyloidogenicity of these proteins is due to mutations in the corresponding protein-encoding genes, and amyloids are hypothesized to be pathogenic and cause toxicity via the loss-of-function effect, which can be either direct or mediated by sequestration of other proteins.

II. Prions, Amyloids, and mRNA Turnover

Several recent studies find that amyloidogenic proteins are widespread in the ribonucleoprotein (RNP) granules found in eukaryotic cells, from yeast to mammals.^106-109 These granules contain non-translating mRNAs and are involved in the control of all stages of mRNA turnover, including storage and degradation. The structure of these granules is highly dynamic, as their protein and RNA composition undergoes significant changes depending on the particular needs of the cell. Two major types of RNP granules, stress granules and processing bodies (P-bodies), are formed in most cells. Stress granules are absent in normal physiological conditions, but assemble rapidly upon the inhibition of translation initiation, which usually occurs in response to various stresses. After the termination of a corresponding stressful condition, stress granules disassemble (for a review see refs.^109,110). Stress granules comprise a number of proteins, many of which contain CBRs / LCRs and are able to form hydrogels containing amyloid-like aggregates in a cell-free system.¹⁰⁸ One of these proteins, Tia1, carries three RNA-binding domains in the N-terminal part and a Q-rich CBR in the C-terminal part. The RNA-binding domains of Tia1 recognize the U- and A-U-rich motifs in the 3′UTRs of mRNAs recruited into stress granules,⁷⁸ and the Q-rich region is required for the recruitment of Tia1 into stress granules, and alone forms protease-resistant polymers in the Hsp70-dependent manner in vivo.^111,112 In vitro, Tia1 forms fibers that, according to Congo Red and Thioflavine-T tests, as well as EM and X-ray analysis, have amyloid structure.^45,113 Moreover, a recent study in yeast demonstrated that full-length mouse Tia1 forms heritable SDS-resistant prion-like cytoplasmic aggregates that co-localize with P-bodies / stress-granules.⁴⁵ Pub1, a yeast ortholog of Tia1, also forms detergent-resistant oligomers and visible prion-like aggregates that co-localize with P-bodies / stress-granules and, like most yeast prions, require the Hsp104 chaperone for their formation.^45,114

The second type of ubiquitous RNP granules is processing bodies (P-bodies). P-bodies interact with stress granules and promote their assembly.^115,116 In contrast to stress granules, P-bodies are constitutively present in cells, although their number and size are increased in response to stress.¹¹⁵ P-bodies encompass untranslated mRNAs and a complex of enzymes involved in mRNA decapping, deadenylation and degradation. In addition to mRNA degradation, P-bodies act as dynamic sites for different stages of mRNA processing, including translation repression and storage.¹¹⁷ In yeast, several P-body proteins contain Q/N-rich regions that can aggregate in vivo.^118,119 In mammals, P-bodies contain fibrils of TNRC6A (GW182)¹²⁰ and DDX6 (RCK, p54)¹²¹ proteins; both of which also have CBRs / LCRs.^120,121

Even though the presence of multiple amyloidogenic proteins with CBRs / LCRs in P-bodies and stress granules led to the hypothesis that functional amyloids are implicated in the biogenesis of these RNPs,^{45,111,118,119} so far there is no proof that these proteins are present in RNPs in the amyloid state.¹²² Indeed, it appears that initial formation of these RNPs is enthalpy driven, depends on multivalent interactions involving both CBRs/LCRs and RNA-binding domains of proteins and RNAs, and leads to the formation of large RNP complexes in extremely dynamic phase-separated liquid-like droplets. Only at subsequent stages these droplets maturate into more stable structures that are more likely to incorporate functional amyloid.^122-124 In the framework of this model formation of amyloid might also manifest an alternative, strictly pathogenic pathway.^122,123 Consequently, the lack of negative selection toward the formation of amyloid by wild type or single-mutation proteins is explained either by the functionality of the amyloid state in the granules or exclusively by the need to maintain CBRs / LCRs prone for multivalent interactions.

A variety of aggregation-prone proteins associated with RNP-granules have been shown to form amyloid aggregates as a result of mutations. In humans these aggregates are hallmarks of multiple diseases. For example, a mutation in the Tia1 protein has recently been linked to Welander distal myopathy.¹²⁵ Also, for TDP-43, which is involved in different stages of mRNA metabolism and reversibly incorporates into stress granules via direct interactions with some RNAs and proteins including Tia1,¹²⁶ mutations, most of which are located in the Q/N-rich LCR, lead to the formation of irreversible intracellular inclusions.¹²⁷ These inclusions are associated with several neurodegenerative disorders, such as frontotemporal lobar dementia and amyotrophic lateral sclerosis¹²⁸ and, in some cases, possess amyloid-like properties.^129,130 Also, wild type TDP-43 is a major component of inclusions in sporadic amyotrophic lateral sclerosis. Another example of pathologic inclusions associated with a stress granule-associated protein are formed by the RNA-binding protein FUS. Mutations in the FUS-encoding gene lead to the formation of FUS inclusions in amyotrophic lateral sclerosis¹³¹ and are also observed in diseases caused by the expansion of glutamine-encoding repeats in other proteins.¹³² Whether these inclusions are related to amyloidogenesis in vivo is still unclear,¹³³ although FUS is capable of forming amyloid-like polymers in vitro.¹⁰⁸ One of interesting hypotheses for both TDP-43- and FUS-associated amyotrophic lateral sclerosis proposes that stress granules are the nucleation sites for the pathologic aggregation of these proteins.^134,135 Alternatively, pathologic aggregation of TDP-43 and FUS may represent off-pathway events in the formation of stress granules, essentially a loss of both nuclear and cytoplasmic functions for TDP-43 and FUS.¹³⁵ Yet another example of disease-associated aggregation involves the mutations in hnRNPA2B1 and hnRNPA1 RNA-binding proteins that lead to formation of multisystem proteinopathy-associated inclusions with fibrillary properties.¹³⁶ Disease-linked mutations in the prion-like domains of hnRNPA1 and hnRNPA2/B1 make possible the formation of a steric zipper, which produces self-complementary β-sheets that comprise the spine of amyloid fibrils accumulating in non-RNP inclusions. This diverts these proteins from physiological folding trajectories connected with the RNP granule assembly and accelerates hnRNPA1 and hnRNPA2/B1 misfolding.^136,137 A low-complexity region of Nab3, a yeast hnRNP homolog, is prone to form amyloid filaments in vitro.¹³⁸ Overall, the RNP granules are highly dynamic structures comprising a repertoire of proteins with low-complexity domains that may either participate in the life cycle of these granules through promiscuous interactions or amyloid formation, or form pathogenic mutation-linked amyloid-like inclusions.

III. Translation-Coupled Prions and Amyloids

Undoubtedly, the best-studied prion in S. cerevisiae is [PSI⁺], which is formed by the translation release factor eRF3 (Sup35). Initially discovered as the non-chromosomal allosuppressor of super-suppressor tRNAs,³⁵ this determinant was proposed³⁶ and proven to be a prion form of Sup35 in a series of studies (for reviews see refs.^139-141). Currently, Sup35 is known to be implicated in a number of functions,¹⁴⁰ some of which are prion-associated. The C-terminal domain of Sup35 acts as a translational release factor eRF3,^142-144 while the N-terminal domain is prion-forming^145-148 and capable of assembling into bona fide amyloid fibrils.^149,150 Interestingly, the Q/N-rich N-terminal prion-forming domain of Sup35 modulates mRNA decay via the regulation of deadenylation,¹⁵¹ although the influence of the [PSI⁺] prion on the degradation of mRNA was not shown. [PSI⁺] not only causes translational readthrough, but also acts as frameshift suppressor, thereby modulating the cellular content of polyamines.¹⁵² The appearance of [PSI⁺] is enhanced by different stressful conditions,¹⁵³ and by the presence of another yeast prion, [PIN⁺],^38,154 which is relatively widespread in yeast populations.^155,156 Interestingly, [PSI⁺] exists in a number of dynamically changing variants,¹⁵⁷ among which there are beneficial,¹⁵⁶ neutral, and harmful⁸⁸ in particular conditions. Moreover, the composition and the size of the [PSI⁺] aggregates, the characteristic feature of [PSI⁺] variants, determines to what extent [PSI⁺]-associated Sup35 retains its functionality.¹⁵⁸

Also, Sup35 interacts with Pub1, an mRNA-binding protein, and this interaction occurs via the Q/N-rich LCRs present in both proteins.⁴⁵ The tubulin-associated protein complex containing Sup35 and Pub1 in oligomeric SDS-resistant prion-like states normally exists in yeast cells and is involved in the maintenance of the integrity of the microtubular cytoskeleton. This complex contains tubulin mRNA and components of the translation machinery, which suggests that it is likely implicated in local cytoskeleton-associated protein synthesis.⁴⁵ Thus, Sup35 can exist in the cell in at least three different states: as a functional monomeric translation termination factor eRF3, monoprotein prion [PSI⁺], which apparently acts as an epigenetic phenotypic modulator through a bet-hedging mechanism, and a functional two-protein [PUB1 / SUP35] prion-like assemblysome. Furthermore, Pub1 is also associated with two self-propagating structures, the one associated with P-bodies / stress granules and not involving Sup35, and the [PUB1 / SUP35] structure.

Remarkable examples of the translation-coupled functional amyloids in multicellular organisms are neuron-specific forms of the cytoplasmic polyadenylation element binding protein (CPEB), ApCPEB, Orb2, and CPEB3 in the mollusk Aplysia californica,^24,159,160 fruit fly Drosophila melanogaster,¹⁶¹ and mouse,^59,162,163 correspondingly. CPEB represents a family of RNA-binding proteins that bind U-rich sequences called CPE elements, which are located in the 3′ untranslated regions of a number of cellular mRNAs. Such binding either promotes, or inhibits translation.¹⁶⁴ Neuron-specific isoforms of CPEB have been hypothesized to provide local protein synthesis in the synapses, which is important for the formation of the so-called “synaptic tag” that stabilizes long-term functional and structural changes in the synapse.^{24,159,165-167}

Current models for ApCPEB postulate that in a naïve synapse it persists in a monomeric form and acts as a repressor of translation. Stimulation of the neurons with serotonin leads to the formation of multimers of ApCPEB. Such induction of the formation of the prion state in response to a physiologically relevant stimulus represents a key feature of a functional prion. Furthermore, in neurons, multimers of ApCPEB self-propagate and exist in the same physiological conditions as the monomers. The ApCPEB multimers possess amyloid properties and are, in fact, an active form of ApCPEB. Indeed, their binding with multimer-specific antibodies destabilizes the maintenance of long-term facilitation.²⁴ Moreover, studies in yeast revealed that an amyloid isoform of ApCPEB possesses multiple hallmarks of a bona fide prion.^160,168

Like ApCPEB, Orb2 exists in two states in the brain of D. melanogaster, monomeric and amyloid-like oligomeric. Multimerization of Orb2 is induced following stimulation with dopamine, octopamine, or tyramine, and persists up to 24 h.¹⁶¹ The Orb2 locus encodes six proteins, only two of which, Orb2A and Orb2B, are CPEB orthologs containing the same prion domain. Orb2A is shorter and, being fused to GFP, forms fluorescent foci, while Orb2B does not form aggregates alone. Endogenous oligomers consist of both, Orb2A and Orb2B, although Orb2A is critical for the involvement of Orb2B in the aggregated state. Mutational inactivation of Orb2A oligomerization leads to the impairment of long-term memory.¹⁶¹ Moreover, Orb2A stability and oligomerization was observed to be controlled by the protein network consisting of Lim kinase, protein phosphatase 2A, and Tob transcription regulator.¹⁶⁹

Finally, three recent manuscripts analyzed aggregation and self-perpetuation of the mouse ortholog of ApCPEB, CPEB3.^59,162,163 Studies of an ectopically expressed purified protein confirmed the ability of CPEB3 to form typical amyloid fibers with a characteristic birefringence upon Congo Red staining. Studies in yeast established that CPEB3 can act as a bona fide prion, i.e. a heritable protein conformation. Also, studies in yeast and mice uncovered some details of CPEB3 prionization: both CPEB3 expression and its aggregation are promoted by synaptic stimulation and occur upon de-SUMOylation of the CPEB3 protein and through its interaction with the actin cytoskeleton. Noteworthy, in agreement with the hypothesis that the CPEB3 prion acts as a synaptic tag in the maintenance of long-term memory persistence, CPEB3 aggregation not only self-perpetuates, but also supports the translation of proteins essential for the functioning of the synapse, such as GluR receptors, as well as proteins essential for CPEB3 regulation, such as actin and SUMO.^59,162,163 Overall, CPEB proteins form functional amyloids with prion-like properties that control long-term memory and possess a complex regulatory network consisting of chemical and protein regulators.

FROM INDIVIDUAL AMYLOIDS TO FUNCTIONAL AMYLOID NETWORKS: PIECES OF A PUZZLE

The diversity of amyloids formed by RNA-binding proteins and proteins involved in mRNA processing and regulation identified to date suggests that amyloid formation is an important component of the key cellular processes related to all stages of mRNA metabolism. In the field of RNA regulation, amyloid-forming proteins participate in transcription, pre-mRNA splicing, mRNA transportation, storage, translation and degradation (Fig. 1). The feature that unites all RNA-modulating amyloids is the presence of CBRs / LCRs. Indeed, all proteins listed in Table 1 have such regions, and in 19 of 22 listed proteins they are rich in Q and/or N residues (Table 1, right column). This is not surprising, as the fact that Q/N-rich CBRs / LCRs contribute to amyloid formation is known for a long time.^51,53 The most representative group illustrating this compositional bias are yeast prions, almost all of which are Q/N-rich. In addition, the Q/N domains of approximately 50 yeast proteins form detergent-resistant aggregates in vivo when overproduced, while 17 of them demonstrate prion-like properties.¹⁴

A number of studies also demonstrated that Q/N-enrichment is a common feature of amyloids in other organisms, including humans. The first group of such amyloids is represented by pathological amyloids, which are formed due to mutations. Such mutations can either be single-residue substitutions (as in the case of FUS, TDP-43, and HNRNPA) or occur through recombination-based mechanisms leading to the expansion of poly-Q-encoding repeats (huntingtin, atropnin-1, etc.). The second group includes functional amyloids, for which aggregation is unrelated to mutations and begins in response to specific stimuli (e.g. synaptic stimulation), or is associated with the assembly of specific intracellular structures (e.g., microtubular cytoskeleton or, possibly, stress granules). Most proteins forming the Q/N-rich amyloids and, in the case of functional amyloids, the amyloids themselves, are functionally related to the metabolism of RNA. Moreover, analysis of the functions of the proteomic subset of Q/N-rich proteins in S. cerevisiae demonstrates that functional groups overrepresented in this subset in comparison with the entire proteome, are related to transcription and RNA-binding (Table 2). This poses an important question: why does Q/N-enrichment strongly correlate with RNA-modulating activities?

TABLE 2.

Functional categories^* overrepresented in the subset of Q/N-rich proteins^** in comparison with the entire proteome of S. cerevisiae

Function	Total number of proteins in category	Number of Q/N-rich proteins	The level of significance
GO:0004674 protein serine threonine kinase activity	128	29	0.05
GO:0004713 protein tyrosine kinase activity	8	4	0.028455
GO:0016251 general RNA polymerase II transcription factor activity	72	20	0.05
GO:0005515 protein binding	486	62	0.04775
GO:0003702 RNA polymerase II transcription factor activity	147	50	0.05
GO:0030276 clathrin binding	13	6	0.047577
GO:0016563 transcription activator activity	50	26	0.05
GO:0016564 transcription repressor activity	39	16	0.05
GO:0004672 protein kinase activity	134	30	0.05
GO:0003723 RNA binding	442	56	0.04131
GO:0046914 transition metal ion binding	398	62	0.05
GO:0019899 enzyme binding	38	9	0.026086
GO:0016773 phosphotransferase activity alcohol group as acceptor	196	34	0.05
GO:0016566 specific transcriptional repressor activity	21	7	0.0416
GO:0001071 nucleic acid binding transcription factor activity	137	50	0.05
GO:0019208 phosphatase regulator activity	26	10	0.05
GO:0019888 protein phosphatase regulator activity	26	10	0.05
GO:0016301 kinase activity	222	34	0.047667
GO:0003704 specific RNA polymerase II transcription factor activity	56	24	0.05
GO:0019789 SUMO ligase activity	5	3	0.013763
GO:0030234 enzyme regulator activity	256	41	0.05
GO:0035091 phosphoinositide binding	66	16	0.048875
GO:0004535 poly(A)-specific ribonuclease activity	5	3	0.013763
GO:0008289 lipid binding	95	23	0.05
GO:0030528 transcription regulator activity	279	88	0.05
GO:0043565 sequence-specific DNA binding	261	77	0.05
GO:0008143 poly(A) RNA binding	7	4	0.038469
GO:0008270 zinc ion binding	313	56	0.05
GO:0060589 nucleoside-triphosphatase regulator activity	131	20	0.006615
GO:0070717 poly-purine tract binding	7	4	0.038469
GO:0008601 protein phosphatase type 2A regulator activity	5	3	0.013763
GO:0003729 mRNA binding	52	22	0.05
GO:0003727 single-stranded RNA binding	15	7	0.0482
GO:0016455 RNA polymerase II transcription mediator activity	30	8	0.029088
GO:0003700 sequence-specific DNA binding transcription factor activity	137	50	0.05
GO:0005543 phospholipid binding	84	21	0.05

^*Functional analysis of the subset was performed in the “Gene Ontology” Database (http://www.geneontology.org/); fractions of proteins were compared by “GoMiner” software.

^**The selection of Q/N-rich proteins from the S. cerevisiae proteome was obtained using the SARP algorithm.¹⁷⁹ The probability thresholds were set to 10⁻⁶ for single residue CBRs and 10⁻¹² for multiple residue CBRs; CBRs were detected for the N and Q residues, as well as for the Q/N pair.

First, poly-Q tracts have their own transcriptional activities. The incorporation of poly-Q stretches into the sequences of transcription factors stimulates their activity in different systems.¹⁷⁰ Correlatively, the expression of poly-Q tracts in yeast causes changes in the transcriptome, and, although these changes depend on both the length of poly-Q tract and presence of the nuclear localization signal (NLS), they occur even in the case of relatively short stretches of 23Qs without NLS in a protein, which is unable to aggregate.¹⁷¹ The same is partly true for poly-N, since 104N fused with GFP and directed by NLS causes repression of transcription from PHO84 and HSP104 promoters.¹⁷² Additionally, analysis of the 104N aggregates¹⁷² supports the hypothesis that poly-Q (or poly-N) aggregates sequester Q/N-rich transcription factors thus modulating the transcriptome profile (see section I).^94,173 Although this hypothesis has been experimentally proven,^94,173 the real contribution of such sequestration to the transcriptomic effects of Q/N-rich amyloids remains unclear.

The recruitment of transcription factors by poly-Q aggregates manifests the influence of Q/N-rich regions on protein-protein interactions. In general, the presence of CBRs / LCRs in protein sequences might affect the probability of their interactions.⁵² Computational analysis of the Q/N-rich proteins of S. cerevisiae revealed that Q/N-rich proteins generally tend not to interact with each other.¹⁷⁴ However, known yeast prion-forming proteins conversely demonstrate a significant tendency to interact with other Q/N-rich proteins.¹⁷⁴ We propose that such selection against interaction of Q/N-rich proteins might have occurred through evolution to prevent non-specific binding between Q/N-rich regions, which is likely to lead to the co-aggregation of proteins and loss-of-function, which occurs in the case of various amyloid-associated diseases.

Simultaneously, similar Q/N-rich regions are widespread in subunits of different protein complexes. While the proteome of S. cerevisiae contains more than 400 protein complexes,¹⁷⁵ approximately 40 of them possess one, and 10 – several Q/N-rich subunits (Fig. 2). All complexes with more than one Q/N-rich subunit are involved in mRNA metabolism (transcription, storage, and nucleocytoplasmic transport). Such enrichment with multiple Q/N-rich subunits in protein complexes suggests that Q/N-rich regions could play a role in the assembly of these complexes. Interestingly, the prion-forming proteins Cyc8, Swi1, Sup35, Pub1 and Nup100 are the components of large protein complexes. Currently, except for Pub1 / Sup35, for which interaction through the Q/N-rich regions is essential for the formation of a tubulin associate complex implicated in the maintenance of microtubule integrity,⁴⁵ for most of these proteins it is unclear whether or not their prionization destabilizes or promotes the corresponding protein complexes. However, with the exception of a subunit of nuclear pore complex Nup100, the prionization of these proteins has specific phenotypes. In the case of Nup100, prion can be detected only biochemically or using the Nup100-YFP reporter and has no phenotypic manifestation, although [NUP100⁺] aggregates sequester other Q/N rich Nup proteins (at least, Nup116 and Nup145).⁴³

FIGURE 2.

Q/N-rich subunits of protein complexes in S. cerevisiae and their interactions. Shown is the interactome of Q/N-rich subunits of protein complexes containing more than one Q/N-rich subunit. Lines indicate physical protein-protein interactions (according to data from “String” database, http://string-db.org/). Experimentally proven prion and amyloid-forming proteins are indicated by black circles, while other Q/N-rich proteins are indicated by gray circles. Light gray rectangles indicate corresponding protein complexes (including their names, as indicated).

Combining the data from the aforementioned studies, we propose that the interactome of prion- and amyloid-forming proteins includes specific groups of interactions occurring due to the presence of CBRs / LCRs in the amino acid sequences of these proteins. We may highlight 3 groups of such specific interactions: (i) prion or amyloid with other prions and amyloids, (ii) subunits of protein complex with a prion- or amyloid-forming protein in the case that this protein is a subunit of a complex, and (iii) interaction between corresponding prion or amyloid with LCR-containing proteins. Combined, this network of interactions is likely to make a significant contribution to the manifestation of the corresponding amyloid. Unfortunately, current data on proteomics of amyloids (or amyloidomics) is very poor, and verification of this hypothesis will only be possible in the future studies. However, several methods have been recently developed allowing to screen proteomes for amyloids. Such methods, TAPI¹⁷⁶ and PSIA,¹⁷⁷ facilitate rapid identification of amyloid proteins. PSIA uses two-dimensional difference gel electrophoresis (2D-DIGE) for the separation of proteins. The limitation of this method is that it does not allow the identification of proteins with extreme isoelectric points.¹⁷⁷ TAPI employs very efficient high-performance liquid chromatography to separate proteins, but utilizes resistance to sodium dodecyl sulfate, to which not all amyloids are resistant, as a screening criterion.^114,176 Despite the limitations, these methods were validated by detection of a wide spectrum of known amyloids and are promising for the identification of novel amyloid-forming and amyloid-associated proteins in different organisms at the proteomic level.

To conclude, RNA-modulating amyloids exhibit an extremely wide diversity of biological roles, including transcription, translation, storage and degradation of RNA, as well as pathogenesis. The ability of a protein to form amyloids and its RNA-modulating activity are closely linked in certain cases, as they are determined by the same sequences rich in Q and N. Such Q/N-rich sequences possess multiple activities: (i) the formation of amyloids, (ii) DNA- and RNA- binding, (iii) the participation in the protein-protein interactions, and likely (iv) the formation of protein complexes. Such a diversity of activities engenders diversity of functions and manifestations of Q/N rich amyloids. Simultaneously, the question is: what is the real number of amyloids in proteomes of different organisms? The amyloids identified to date are likely to be only pieces of a puzzle, since they were identified only by studies addressing particular proteins, and their real prevalence in proteomes could be considerably higher. Further proteomic studies of amyloids should highlight these issues and explain the biological significance and roles of amyloid structure.

ABBREVIATIONS

CBR

compositionally biased region in protein sequence

eRF

eukaryotic translation release factor

hnRNP

heterogeneous nuclear ribonucleoprotein

Hsp

heat shock protein

LCR

low complexity region in protein sequence

NLS

nuclear localization signal

PSIA

proteomic screening and identification of amyloids

RNP

ribonucleoprotein particle

TAPI

technique for amyloid purification and identification

YFP and GFP

yellow and green fluorescent proteins, respectively

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

No potential conflicts of interest were disclosed.

Funding

This work was supported by National Institutes of Health (grant 7 R01 GM070934-06 to ILD), by the grant of the President of the Russian Federation (MK-4854.2015.4 to AAN and KSA), by Russian Foundation for Basic Research (grants 16-34-60153 to AAN and 16-34-00582 to SAB, KSA, and AAN), and St. Petersburg Government (to AAN and KSA). The authors acknowledge Saint-Petersburg University for a research grants 1.50.2543.2013 (to AAN and SGI), 15.61.2218.2013 (to SAB), and 1.37.291.2015 (to KSA, SAB, and AAN).