REVIEW ARTICLE Prion-like proteins: from computational approaches to proteome-wide analysis Marcos Gil-Garcia, Valentın Iglesias s and Salvador Ventura , Irantzu Pallare noma de Barcelona, Spain Departament de Bioquımica i Biologia Molecular, Institut de Biotecnologia i de Biomedicina, Universitat Auto Keywords Prions are self-perpetuating proteins able to switch between a soluble state prion; prion-like protein; functional amyloids; and an aggregated-and-transmissible conformation. These proteinaceous prion-like prediction; bioinformatics; entities have been widely studied in yeast, where they are involved in hered- proteome screenings itable phenotypic adaptations. The notion that such proteins could play Correspondence functional roles and be positively selected by evolution has triggered the s, Departament de S. Ventura and I. Pallare development of computational tools to identify prion-like proteins in dif- Bioquımica i Biologia Molecular, Institut de ferent kingdoms of life. These algorithms have succeeded in screening mul- Biotecnologia i de Biomedicina, Universitat tiple proteomes, allowing the identification of prion-like proteins in a Autonoma de Barcelona, Bellaterra, diversity of unrelated organisms, evidencing that the prion phenomenon is Barcelona 08193, Spain well conserved among species. Interestingly enough, prion-like proteins are E-mails: salvador.ventura@uab.cat (SV); not only connected with the formation of functional membraneless pro- irantzu.pallares@uab.cat (IP) Tel: +34 935868956 (SV); +34 935812154 (IP) tein–nucleic acid coacervates, but are also linked to human diseases. This review addresses state-of-the-art computational approaches to identify Marcos Gil-Garcia and Valentın Iglesias prion-like proteins, describes proteome-wide analysis efforts, discusses these contributed equally to this article unique proteins’ functional role, and illustrates recently validated examples in different domains of life. (Received 3 February 2021, revised 7 May 2021, accepted 28 May 2021) doi:10.1002/2211-5463.13213 Edited by Irene Dı́az-Moreno In recent decades, our perception of the prion phenom- [5,6]. Most of our knowledge on these polypeptides’ ena has evolved remarkably. Initially considered strictly self-replicating properties comes from the study of as infectious proteinaceous elements linked to mam- Saccharomyces cerevisiae prion proteins, where the malian neurological disorders, prions and prion-like transition from a soluble conformation to an aggre- proteins are now known to play critical roles in regulat- gated state has been related to heritable phenotypic ing evolutionary conserved biological processes [1,2]. changes [7-10]. In most yeast prions, but not all, the Prions are a particular subset of proteins able to prion activity is encoded in a fraction of the protein, adopt different conformational and functional states, the prion domains (PrD) that share a set of composi- where at least one of them is amyloidogenic and tional features since they correspond to low- transmissible, templating homologous polypeptide complexity regions (LCRs), with enrichment in glu- sequences, even across species [3,4]. These biomole- tamine (Q) and asparagine (N) and depletion in cules show the ability to aggregate into multiple differ- charged and hydrophobic residues. They also share ent amyloid structures, that is, strains, with distinct conformational traits as they are embedded in long physicochemical properties and propagative features intrinsically disordered regions (IDRs). These PrDs are Abbreviations IDR, intrinsically disordered region; LCR, low-complexity region; LLPS, liquid-liquid phase separation; PrD, prion domain; PrLD, prion-like domain; RBD, RNA-binding domain; RNP granules, ribonucleoprotein granules. 2400 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes both necessary and sufficient for prion conformational of the challenging PrLDs biology would benefit from conversion [11-14]. the global view provided by large scale analysis. We Classical amyloid prediction algorithms are blind to describe a collection of computational approaches that the amyloidogenic potential of Q/N-rich stretches, sug- exhibit significant distinctive features, despite being all gesting that the sequence-specific physicochemical prop- trained on yeast prions sequences. The application of erties that drive the conformational promiscuity of these algorithms to proteome-wide analyses has led to prions do not conform to the standard features that identifying relevant PrLDs in organisms of divergent govern the formation of canonical amyloid aggregates biological complexity. We discuss their involvement in (e.g., high hydrophobicity and intrinsic b-sheet propen- cellular regulatory functions, their role as molecular sity). The extensive research devoted to the characteriza- tools for adaptation, and their linkage to disease. tion of the PrDs of the yeast prions Sup35 and Ure2p Overall, predictors have allowed the discovery of (their prionogenic states are known as [PSI+] and PrLDs functions that would have never been unveiled [URE3], respectively) has enabled to dissect the different analyzing individual proteins. molecular determinants behind prion conversion [15- 18]. Collectively, this knowledge has fueled the develop- Computational tools for the prediction ment of a collection of bioinformatics tools to identify of prion-like domains proteins bearing similar prion-like domains (PrLDs) in biological relevant sequence databases as well as in com- In recent years, and especially since next-generation plete proteomes [13,14,19,20,21,22,23,24,25,26,27]. sequencing techniques became available, annotated The implementation of computational proteome- sequences on public databases have exponentially wide analyses has identified thousands of new proteins increased. In this context, large-scale bioinformatics potentially harboring PrLDs in a wide variety of analysis should allow screening for prion-like proteins organisms, ranging from viruses to humans across organisms of distant clades. However, conven- [13,21,28,29,30,31,32,33,34,35]. Thus, supporting the tional aggregation prediction algorithms often fail to idea that prions, and their evolutionary conserved discriminate prions from nonprions correctly [27,46,47] capacity to experiment conformational transitions, are (Table 1). exploited by cells as important functional elements. Therefore, there was a need for new programs that Although the prionic load varies significantly among exploit the specific features of PrD in their predictions. organisms, accounting for < 1% of human proteins or Two different models were proposed for decoding the up to 10% in Plasmodium falciparum [31,33,36], they forces driving conformational changes from primary seem to conserve functional and structural traits. Col- sequences. The compositional model considered that it lectively, prion-like subproteomes consist of modular would be necessary a large number of weak interac- proteins containing multiple functionally associated tions distributed along the PrD for which a biased domains, with a particular abundance of RNA- and amino acidic composition was required [27]. The soft- DNA-binding motifs [37,38]. In this context, it is amyloid model relied on the presence of stretches of becoming increasingly evident that PrLDs are involved mild amyloid potential, where the amyloid load is in highly connected interaction networks and associ- more diffuse than in ‘classical’ pathogenic amyloids, ated with the formation of dynamic membraneless embedded within an intrinsically disordered PrD intracellular compartments through liquid-liquid phase [26,48]. Several groups have approached prion-like pre- separation (LLPS) [39-41]. Indeed, the presence of diction applying these different, but as we will see, structurally disordered regions in proteins correlates complementary perspectives, resulting in multiple algo- with the propensity to establish weak–transient interac- rithms (Table 2) [13,14,19,20,21,22,23,24,25,26,27]. tions that drive liquid-to-solid phase transitions This section will review six representative state-of-the- [38,42]. Remarkably, in the human proteome, the art bioinformatics tools, which have had a significant PrLD subset includes well-characterized prion-like pro- impact on identifying novel prion-like domains and teins such as FUS, TDP-43, TIA1, or hnRNPA1. their associated proteins. Mutations in these proteins are associated with neu- rodegenerative disorders, and they often map at their Composition-based predictors PrLDs, promoting the conversion of the condensate state into solid amyloid-like aggregates [43-45]. Several lines of evidence, including the presence of Q/ Despite significant progress in our knowledge of the N-rich domains in yeast prions that form amyloids prion phenomenon, both in beneficial cellular pro- (and their inactivation upon mutation of Q and N resi- cesses and pathological conditions, our understanding dues) [49], Q-repeats in aggregating proteins involved FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2401 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. Table 1. Performance of aggregation prediction methods when identifying prionic sequences. AGGRESCAN [132], PATH [133], RFAmyloid [134] and AmyloGram [135] prediction methods were run with default parameters and results were obtained using their standard thresholds. PrionW [19] is intended for predicting prion-like proteins and was used for performance comparison (in italics). The sensitivity, specificity, precision, accuracy, Matthews correlation coefficient (MCC) and F1 Score were calculated from yeast prion domains and prion- like domains experimentally validated by Alberti and coworkers [13]. The authors characterized the domains for their amyloid and prion forming ability in four assays and scored them from 0 to 10. As described previously [26], those domains that were positive in all four assays and scored ≥ 9 were considered prions, and nonprions those sequences scoring ≤ 2 and being positive in one assay at maximum. The dataset was composed of 12 true positives (TP) (including bona fide prions Sup35, New1, Swi1, Ure2p, Rnq1) and 39 true negatives (TN). False negatives and false positives are abbreviated as FN and FP, respectively. Algorithm TP TN FP FN Sensitivity Specificity Precision Accuracy MCC F1 score PrionW 11 37 2 1 0.92 0.95 0.85 0.94 0.84 0.88 AGGRESCAN 0 39 0 12 0.00 1.00 – 0.76 – 0.00 PATH 3 39 0 9 0.25 1.00 1.00 0.82 0.45 0.40 RFAmyloid 8 15 24 4 0.67 0.38 0.25 0.45 0.04 0.36 AmyloGram 10 11 28 2 0.83 0.28 0.26 0.41 0.11 0.40 in human disorders [25], or the experimental demon- proteins; in an attempt to understand how protein stration that inter-chain hydrogen bonding between variation could modulate the prion-like outcome [52], glutamines stabilizes antiparallel b-strand conforma- an effect previously demonstrated for the human tions [50], indicated that these amide-carrying amino hnRNPDL prion-like protein [53]. acids played an important role in the self-propagation Lindquist and coworkers applied another strategy to of aggregated species [25,49]. These observations lie unravel the determinants behind prion conversion. behind the compositional model for prion-like conver- They trained a hidden Markov Model (HMM) on PrD sion, according to which many weak, sequence- detection using the four bona fide yeast prions known independent interactions across a sizeable low com- at that time and ranked with it the entire yeast gen- plexity region would drive the structural conversion ome. The highest 100 predicted prion domains were into an amyloid aggregate [27]. Different algorithms tested for aggregation, amyloidogenicity, and the abil- have been developed following this principle. They ity to maintain the prion phenotype on a chimeric essentially differ in the contribution assigned to each Sup35, which allowed them to discover a new real of the 20 proteinogenic amino acids. prion, Mot3. The 28 domains, which scored highest in The identification of a scrambled version of Sup35 those experiments, were later used to feed the HMM. yeast prion that undergoes prion conversion without This tool was made available as the prion-like amino overexpression allowed Toombs et al. to design a strat- acid composition (PLAAC) prediction algorithm in the egy to quantify each amino acid’s influence on this form of a web server and a java code. PLAAC incor- process. They mutated an 8-amino acid segment of this porates the possibility of adjusting the HMM calcula- Sup35 version randomly and sequenced those clones tions with either a set of prepopulated species that kept the parental prionic behavior. The prion background or the entire input sequences. The algo- propensity was then defined as the log-odds ratio of rithm also incorporates PAPA, DIANA, and FoldIn- each amino acid frequency among the prion-forming dex calculations and returns their outputs. clones, relative to the starting library [14]. The method Afsar Minhas, Ben-Hur and Ross further worked better on top of predefined disordered regions approached the phenomena from a machine-learning (for which they applied the FoldIndex algorithm [51]) perspective. Two main barriers hamper the use of neu- and smoothing the scores using 41-residue windows. ral networks to develop a binary prion classifier: The resulting algorithm was made available through a (a) the low number of annotated bona fide PrDs and web server and a python code as Prion Aggregation (b) the annotated PrDs usually entail an area larger Prediction Algorithm (PAPA) [27] and used to design than the effective sequence required for prion conver- synthetic prions that maintained the Sup35 Q+N sion. They addressed these limitations using a super- content. In a follow-up study, Cascarina and Ross vised learning method known as multiple-instance updated PAPA by modifying the minimal PrLD length learning (MIL), which allows modeling with some and lowering the requirements for positive detection degree of uncertainty, while using sequence composi- in order to study the influence of genetic variation, tion alone as an input. The method, named prion mutations, or the effect of post-transcriptional and RANKing and classification (pRANK), was trained post-translational modifications on putative prion-like on top of 22 known Q/N rich yeast PrDs in the 2402 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes Table 2. Prion-like prediction methods and applied analytical rationale. Algorithm Strategy Brief description/Underlying rationale Availability References DIANA Compositional Identifies Q/N-rich domains by counting Qs and Ns in 80 – [25] consecutive amino acids windows and retrieves the most enriched stretch above a 30-Q/N threshold. Stretch length and minimum Q/Ns content are based on the length and Q/N percentage of Sup35 and Ure2p yeast prions and human disease-causing polyQ-expansions LPS Compositional The program is designed to retrieve any compositionally biased Script [23,136] sequence, and it has been used to identify Q/N-rich regions as a proxy to PrLD. It first searches for all possible single amino acid bias by comparing each window against the inputted background frequencies and retrieving the lowest probability stretch. Subsequent updates allowed automatic calculation of bias for multiple residue types by checking if their combined probability was lower than those of the individual residues separately. LPS calculates for and against biases PAPA Compositional Prediction is made on disordered segments, exploiting an amino Web [14,27] acid propensity scale obtained by random mutating a short server + script stretch of a prionic Sup35 variant PLAAC Compositional Applies an HMM trained on 28 yeast PrD and PrLD with high Web [13,24] experimental prion propensity against user-selected server + script backgrounds of amino acid frequencies pWALTZ Amyloid Applies the WALTZs’ experimentally-derived amyloid propensity Executable [26,137] scoring matrix to longer stretches, averaging the amyloid load over 21-residues stretches and retrieving the strongest amyloid- core PrionW Compositional + Amyloid Disordered fragments with a minimum QN content are Web server [19] evaluated with pWALTZ. The QN-threshold can be adjusted for different species’ background frequencies PrionScan Compositional Uses an unsupervised classifier and a statistical representation Web [21,22] of PrLD relying on the amino acid frequencies of positive and server + Database negative sequences in Lindquist’s dataset. PrionScan incorporates a built-in database that regularly updates its predictions for UniProt KB releases pRANK Compositional/Machine Implements a supervised learning strategy, trained on 22 known Web [20] learning Q/N-rich yeast PrD as positive sequences and the remaining server + script proteome with < 90% similarity to them as negative ones Lindquist’s dataset and released as a web server and a in well-performing prion-like sequences. PrionScan python code [20]. pRANK showed an excellent dis- was applied to screen the whole UniProt KB rendering crimination performance on top of the yeast proteome. notable differences for organisms of different taxo- However, the high weight assigned by pRANK to Q nomic categories; and identifying individual proteomes results in an inherent bias to detect false positives aris- with a high prionic load, such as the ones of Dic- ing from polyQ stretches, something that should be tyostelium discoideum and P. falciparum with 20% and regarded with care when working with strongly com- 10% of predicted sequences, respectively. PrionScan positionally biased proteomes. was made available through a web server, a Perl code, Ventura and Sancho evaluated Lindquist’s top 29 and a built-in database. For updated predictions, PrLD dataset and established their amino acidic fre- PrionScan regularly analyses UniProtKB database quencies. Contrasting their composition with other releases and stores them into its built-in database, cur- high scoring but nonprion Q/N-rich PrLD sequences rently holding > 28 000 prion candidates. Over the from the same study, they built up the predictive years, several groups have applied prediction methods approach PrionScan. This algorithm revealed a posi- to generate online databases for their prion-like candi- tive bias toward Q, N, S, and Y residues and a nega- dates [54,55], but by the time of this writing, only tive one toward charged residues, as well as C and W, PrionScan was currently active. FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2403 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. Soft amyloid-based predictors by comparing the amino acid compositional similarity of protein candidates to the average composition of Scrambled Sup35 and PrDs can adopt prionic states; bona fide yeast prions [13,14,21,22,27,28,29,58,59]. On however, their propensities differ from that of the the other hand, pWALTZ and PrionW algorithms, respective wild-type forms, indicating that sequential which consider in their predictions prion amyloid traits modulate conversion, which is consistent with the propensity, have also shown to be highly accurate for in-register parallel b-sheet structure proposed for their discriminating yeast prions [19,26], and they have been amyloid states [56,57], still conventional amyloid predic- applied not only for the characterization of large data- tors were blind to the sequential determinants of such sets [32,33,60] but also for the discovery of soft- amyloidogenicity. Ventura and co-workers realized that amyloid cores in yeast prions [61]. Further studies these yeast PrDs had a lower amyloid propensity and identified the one in Sup35 as the region required for that it was diluted into more amino acids than in classi- the induction, propagation, and inheritance of the cal amyloids [26]. These made them postulate the soft- prion state in mammalian cells [62]. amyloid model, in which a short linear stretch of softer Regardless of the relative contribution of each of amyloid propensity inside an IDR would kick start the the above-discussed features to prion-like conversion, conformational switch toward the aggregated, amyloid a combination of compositional and soft-amyloid state [26]. They used the position-specific amyloid pre- requirements increases the stringency of the prediction, dictor WALTZ and extended its length to 21 residues thus increasing the confidence in predicted candidates’ based on this being the minimum length of a transmissi- performance (Fig. 1). Remarkably, this combinative ble b-fold at that time (as seen in the atomic structure of strategy led to identifying the transcription terminator HET-s prion from the fungus Podospora anserina). This Rho factor as the first prion protein in bacteria approach, named pWALTZ, was tested on top of Lind- [30,60,63]. quists’ PrLD dataset, showing higher discriminatory Of note, the interplay between compositional bias potential than PAPA. Moreover, it identified amyloid and the presence of short-amyloid stretches has cores in PAPA synthetic prions and disease-related enabled predicting mutations impact on the aggrega- mutations in human prion-like proteins and the lack of tion propensity of prion-like proteins [64,65] in various them in PAPAs’ nonprion engineered sequences [26]. diseases related to mammalian prion-like protein vari- pWALTZ was made available as an executable file. ants [66-69]. As pWALTZ was trained on top of yeast PrD, it was not designed for detecting IDR or PrLD on full protein sequences but relied on users inputting putative prion- Proteome-wide analysis of prion-like like domains. As such, its predictive power was expected proteins in different kingdoms of life to decrease when applied in proteomic scale screenings, The study of prions in baker’s yeast has been pivotal as it could mistakenly detect as soft-amyloids regions to understand prion properties and their associated- belonging to the hydrophobic core of folded proteins or functions [70,71]. The yeast HMM-predicted subpro- membrane-associated proteins. The same group applied teome presented an enrichment in proteins involved in pWALTZ on defined sequence contexts in a novel PrLD gene expression processes (transcription factors and predictor to overcome these limitations. This software, RNA-binding proteins) consistent with their role as which was named PrionW, defines the PrLD by first epigenetic devices of adaptation to stress [13]. applying disorder calculations (by deploying FoldIndex By definition, canonical prions exhibit QN-rich predictions) and secondly calculating Q/N-richness on domains but also the ability to populate amyloid top of those IDRs [19]. It allowed users to specify the supramolecular conformations and a Hsp104-dependent minimum Q/N requirement for their sequences to be cell-to-cell propagation [72,73]. Although it is not the considered prion-like, which addressed species-specific scope of this review, it is worth to mention that certain background compositional bias. PrionW selects the proteins lacking these canonical traits may act as prions, IDR with the highest number of Q/N and passes it to including the so-called mnemons, which can suffer con- pWALTZ, where the highest-scoring soft-amyloid core formational switches, but cannot propagate to daughter is selected. PrionW was made available as a web server. cells, acting as a kind of cellular memory [74,75]. The success in the computational identification of Predicting prion-like proteins in complete previously undescribed canonical yeast prions has proteomes encouraged the discovery of similar proteins (generi- Prediction methods like PLAAC, PAPA, and PrionS- cally termed as prion-like proteins) in a wide variety of can have succeeded in identifying prion-like proteins organisms, ranging from virus to higher eukaryotes. 2404 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes Fig. 1. Bioinformatics has been a valuable tool in the discovery of prion-like proteins. (A) Prion-like prediction methods correctly identify PrD in yeast prion proteins Sup35 and Ure2p. Disordered N-terminal regions shown in cartoon representation are models not derived from structural data. (B) Proposed pipeline for optimizing prion-like proteins discovery. Large scale computational analysis offers a powerful alternative to time-consuming and more expensive experimental approaches. Stepwise sequential restrictions in the selection from a pool of initial candidates increase the success rate for discovering novel prion-like proteins. (a) mPAPA corresponds to the 2020 modified version of PAPA (b) pWALTZ requires a previously defined PrLD; thus pWALTZ soft-amyloid core predictions are only shown for PrionW. FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2405 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. In next sections, we review the most recent proteome- nucleus itself and the ribonucleoprotein complex. Not wide analysis aimed to identify such prion-like pro- surprisingly, the most abundant Pfam cluster associ- teins, delving into their biological functions, and pro- ated with PrLDs corresponded to DNA- and RNA- viding data on experimentally validated examples binding domains such as the RRM domain. Therefore, (Fig. 2). even if the Gitler and Ventura laboratories departed from different premises and approaches, their studies converged to show an association between PrLD and The human proteome RRM moieties. Overall, the predictions were consistent The connection between FUS and TDP-43, two proteins with evidence from the literature that reported a cru- bearing a PrLD together with RNA recognition motifs cial role of human prion-like proteins in forming (RRM), and neurological diseases suggested a connec- ribonucleoprotein (RNP) granules involved in the tion between this kind of architecture and pathology post-transcriptional regulation of gene expression. [37]. RRM is one of the most versatile RNA-binding An analysis of the expression profiles of the 242 pro- domains (RBDs) and is conserved across different king- tein’s in Ventura’s dataset revealed that, in agreement doms of life. Gitler and coworkers used a yeast func- with An and Harrison observations [31], the expression tional screen, together with bioinformatics analysis, on of these polypeptides was not restricted to the nervous top of the RRM-containing human subproteome to system but instead distributed among different tissues, identify new putative prion-like proteins behind neuro- suggesting that, contrary to what was initially thought, logical disorders [76]. They obtained evidence of a they could be involved in diverse functions across the PrLD-enrichment in RNA-binding proteins and human body. For example, the Mediator complex, a revealed the association of yet another RNA-binding transcriptional coactivator required for gene transcrip- protein, TAF15, with neurological damage. tion by RNA polymerase II in eukaryotes [77], con- One successful attempt to computationally identify tains subunits displaying PrLDs in their sequences. PrLD-containing proteins in humans was performed Many of the identified proteins seemed to exert essen- by An and Harrison [31]. In this study, the authors tial functions, which immediately suggested that alter- selected candidates using three different prediction ations in their activities could trigger the development methods (LPS, PLAAC, and PAPA), uncovering that of multiple diseases [78]. they account for approximately 1% of the human pro- Indeed, a very robust association between human teome. A subsequent analysis of the data corroborated prion-like proteins and disease was evident in this a statistically significant relation between human study. Notably, as expected from their broad tissue prion-like proteins and neurological diseases. However, distribution, these polypeptides appeared to be the analysis also indicated that these proteins’ expres- involved in neurological problems, but also in different sion was not exceptionally high in the central nervous types of cancer and viral infections. The reasons system. behind this tight relationship likely rely on (a) the high Ventura and coworkers applied a more stringent connectivity of prion-like protein-protein interactions procedure to identifying human prion-like proteins networks and (b) the inherent aggregation propensity [36]. The prediction scheme consisted of an initial of PrLDs, in such a way that the establishment of screening for the presence of regions displaying com- promiscuous non-native and irreversible intermolecular positional similarity to yeast PrDs using PLAAC, fol- interactions would unfailingly disrupt network fluxes lowed by a refined search of soft amyloid stretches and impact cell viability. using pWALTZ. The collected results revealed a total The best-characterized examples of disease- of 242 proteins harboring PrLDs from an initial data- associated prion-like proteins are the TDP-43 and set of > 70 000 proteins, accounting, in this case, for FUS proteins, as mentioned above [79,80]. TDP-43 0.34% of the human proteome. was identified as the main component of protein aggre- An in-depth analysis of these PrLD-containing pro- gates in patients with amyotrophic lateral sclerosis teins’ biological role evidenced a significant enrichment (ALS) and some types of frontotemporal lobar degen- in RNA and DNA associated processes, particularly eration (FTLD) [81]. TDP-43 possesses two RRMs the ones related to positive regulation of transcription and a C-terminal PrLD, where most of the disease- from RNA polymerase promoter, mRNA splicing, and associated mutations map [82]. In FUS, the PrLD is RNA processing. Analogously, the most enriched located at the N-terminal extreme, and there is only an functions were nucleic acid binding and transcription, RRM [83,84]. Its aggregation is linked to ALS and and the cellular components overrepresented in this FTLD, and mutations in its sequence are involved in subproteome were nuclear structures, including the disease, especially those occurring at its PrLD and in a 2406 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes Fig. 2. Proteome-wide analyses of prion-like proteins in distinct kingdoms of life. Large-scale analyses approximate the content of PrLD- containing proteins in the proteomes unrelated organisms. Experimentally-validated examples (if any) of prion-like proteins and proteome enrichments are indicated for each species. ND, not determined. Created with BioRender.com. C-terminal segment predicted as a nuclear localization body of evidence has demonstrated that such domains signal [37,82]. Other RNA-binding proteins such as are indispensable for forming membraneless compart- EWSR1 and TAF15 are also involved in ALS and ments where RNA and proteins are recruited in FTLD; however, their disease-associated mutations are response to diverse cellular and environmental signals placed outside PrLDs and respond to functional [86,87]. These dynamic structures are formed through a domains inactivation [58,85]. process commonly known as liquid-liquid phase separa- Besides their invariable connection with pathological tion (LLPS) that involves establishing weak and nonco- phenotypes, proteins harboring PrLDs have not been valent transient interactions between prion-like proteins evolutionary purged out. In that context, a growing [41,86,88]. Although the ubiquity of PrLD-containing FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2407 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. proteins in membraneless compartments suggested that ensure proper development and survival. Some of these domains are necessary and sufficient to promote these adaptations seem to be regulated by amyloid LLPS [89,90], recent results have demonstrated that and/or prion-enciphered events [29,101,102]. interactions between residues located at PrLDs and The first detailed biocomputational analysis of RBDs are essential to trigger the formation of such pro- prion-like proteins in plants was conducted by tein condensates in vivo [91]. What is clear is that the Chakrabortee et al. in the Arabidopsis thaliana pro- deregulation of this process can promote an aberrant teome. Applying the compositional algorithm PLAAC, irreversible transition toward a more-solid, aggregated, they identified 474 putative proteins bearing PrLDs, and in some cases, amyloid-like state associated with the accounting for roughly 1% of the proteome [29]. onset of devastating proteinopathies. It seems that both Ontology analysis revealed an enrichment in biological loss and toxic gain of function are at the core of these processes related to gene expression and regulation as pathologies [10,43,45]. LLPS deregulation can be caused well as in RNA-binding. These enriched processes and by changes in protein concentration, evoking the estab- the associated molecular functions are common to lishment of more frequent intermolecular interactions, those reported for the human proteome and, as we will or by point mutations that strengthen LLPS interactions see below, to the ones identified in the rest of analyzed and increase the kinetic barrier for dissociation, altering proteomes, independently of the taxon of origin, some- the material properties of the condensates. The result is, thing that ultimately arises from the fact that prion- in both cases, less dynamic and thus dysfunctional con- like proteins are inherently modular and, as described, densates. In this context, there is a high interest in new PrLDs tend to associate with RNA- or DNA-binding therapeutic approaches that prevent or restore aberrant domains. phase transitions [92,93]. In A. thaliana, the predicted PrLDs were also con- Cells exploit different stratagems to alter the assem- nected to plant-specific biochemical pathways such as bly properties of PrLD-containing proteins. One of those involved in reproductive and flower development these strategies is the process of alternative splicing processes. Four proteins involved in autonomous flow- (AS), a molecular event particularly relevant in human ering exhibited predicted PrLDs, but only the prion-like proteins [94,95]. Batlle et al. [53] have Luminidependens protein (LD) recapitulated some of demonstrated that the self-assembling properties of the the classical characteristics of prion proteins when it hnRNPDL prion-like protein are strongly influenced was expressed and characterized in yeast. by its AS. The three different hnRNPDL-isoforms The LD-PrLD aggregates and can replace the generated by AS clearly differ in their propensity to Sup35-PrLD in yeast promoting the characteristic phe- LLPS, with one forming liquid droplets at very low notypic switch; however, the self-perpetuating confor- concentrations, whereas another one is LLPS- mation corresponds to low molecular weight incompetent. The control of assembly by AS also oligomers, instead of high-molecular-weight amyloid occurs in the prion-like proteins of other organisms, fibrils. Also, contrary to most classical yeast prions, like in Drosophila Orb2/CPEB, where it regulates self- LD-PrLD propagation is not influenced by Hsp104 templating at the synapse, which is correlated with the and might depend instead on Hsp70 or Hsp90 activi- facilitation of long-term memory [96]. ties, a phenomenon described for the noncanonical Not only the final primary sequence of the polypep- fungal prions [GAR+, ESI+ and SMAUG+] [103-106]. tide is relevant for its condensation potential, but this LD protein bears a DNA-binding homeodomain able property is strongly impacted by the post-translational to regulate transcriptional events [107]. It was pro- modification of residues that contribute the most to posed that LD self-assembly could alter its activity the dynamics of condensates, including arginine and evoke chromatin modifications involved in flower- methylation or tyrosine and serine phosphorylation ing decisions, representing an epigenetic memory [97,98]. In this way, hypomethylated FUS has been mechanism previously undescribed in the plant realm. found in insoluble aggregates of FTLD patients, sug- More recently, Jung et al. have identified EARLY gesting that deficient arginine methylation contributes FLOWERING 3 (ELF3), a component of the evening to this pathological phenotype [99,100]. complex related to the plant circadian clock that undergoes prion conversion, functioning as a ther- mosensor in A. thaliana [108,109]. ELF3 harbors a The plant proteome polyQ-PrLD able to transit between a soluble and a Plants cannot move from one place to another the phase-separated state depending on the environmental way animals can. However, they quickly react to their temperature. A. thaliana exploits the self-assembling surroundings and adapt their physiological response to properties of ELF3 as a thermosensory mechanism to 2408 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes promote the plant growth and flowering at specific intrinsically disordered, similar in composition to yeast temperatures. Jung and co-workers revealed that PrDs, and with a cryptic propensity to assemble into ELF3-PrLD length varies across plants species living amyloid-like conformations. This restrictive strategy in different climates, suggesting that this divergence rendered a total of 503 PrLD-containing proteins. This could be the product of an evolutive mechanism for represents 1/3 of those found in composition only the optimization of thermal responsiveness according based predictions; however, they still represent 9.4% to the environmental conditions [110]. This strategy, of the proteome [33], confirming that, effectively, one and in general prion switches, exploit the biophysical of the consequences of the P. falciparum genome bias properties of polypeptides to regulate their conforma- is a prion-like enriched proteome. tional states without requiring a dedicate machinery Functional analysis of this curated subproteome for post-translational modifications, resulting in a provided interesting insights into prion-like proteins’ rapid and intuitive physiological response to tempera- role in this deadly parasite [33]. Again, the most ture fluctuations. enriched biological processes included gene expression Although great strides have been made in the study and transcription regulation, reinforcing a preserved of PrLDs in plants, much remains to be understood functional role of PrLDs across all kingdoms of life. and we envision that significant prion-based molecular More interestingly, the analysis also revealed signifi- processes wait to be discovered in this kingdom. cant associations with specific P. falciparum processes not identified in previous proteome analyses, including protein localization and regulation of vesicle-mediated The Plasmodium falciparum proteome transport. This is important because the traffic of par- The protozoan parasite P. falciparum is the primary asite proteins across the host cell plays a key role in cause of severe malaria and responsible for approxi- host-parasite interactions, pathogenesis, and disease mately half-a-million annual human deaths (World susceptibility [116]. Between the prion-like constituent Malaria Report, November 2020) [111]. Comparative functional domains, the most enriched ones were those genome analyses revealed an exceptionally high AT related to nucleotide-binding, including the ubiquitous content compared to other eukaryotes, with nearly RRM, but also domains not detected in other organ- 80% in coding regions. This is translated into an unu- isms such as the ApiAP2 domain. Indeed, 50% of the sual proteome where the 30% corresponds to LCRs ApiAP2 protein family was shown to bear a PrLD. highly enriched in asparagine residues [112]. Low com- These proteins are the central transcriptional regula- plexity and N-enrichment are two of the characteristics tors in Plasmodium parasites and the other Apicom- of PrLDs, which already suggested that P. falciparum plexa [117], playing crucial roles in gametocyte might be especially rich in prion-like proteins. maturation [118], organ infection [119], and the devel- Singh et al. [112] performed the first computational opment of host immunity [120]. Thus, the analysis survey for PrLDs in P. falciparum, which they defined as uncovered the association of putative PrLDs with consecutive stretches of 80 residues that contain at least essential functions connected with Plasmodium life 30 Q and/or N, unveiling that the 24% of P. falciparum cycle regulation and its adaptation to the host. Addi- proteins harbor at least one of those Q/N-rich domains. tionally, the authors demonstrated experimentally that Considering the inherent aggregation propensity of these PrLDs contained sequences able to sponta- PrLDs, it is intriguing how organisms like P. falciparum neously self-assemble into amyloid-like structures, or D. discoideum can stand with very high prionic loads. which can potentially trigger protein conformational Detailed analyses of their proteomes have revealed an conversion. Although no bona fide prion protein has adapted protein quality machinery (e.g., PfHsp110c and been experimentally validated in P. falciparum yet, SRCP1 proteins in Plasmodium and Dictyostelium, their predicted abundance and the fact that, contrary respectively) to keep their highly aggregation-prone pro- to what was initially thought, they do not correspond teome under control and ensure the fitness these species to spurious sequences but instead seem to exert impor- in their natural environments [113-115]. tant and well-conserved functional roles in the para- To further characterize P. falciparum prion biology, site, might offer a new class of therapeutic targets to Pallares et al. [33] applied a more stringent biocompu- treat malaria. tational approach. First, as in Singh et al., LCR enriched in Q/N residues were identified. Subsequently, The bacterial proteomes these regions were evaluated with the PAPA algo- rithm, and the positive sequences were scanned with Bacteria are single celled microorganisms adapted to pWALTZ. Thus, the predicted PrLDs were Q/N rich, multiple and diverse environments. The evidence that FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2409 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. eukaryotic unicellular organisms exploit prion mecha- essential for the termination of transcription by RNA nisms to adapt in front of environmental fluctuations polymerase [124]. Although the prion activity of C. bo- and the discovery of an increasing number of func- tulinum Rho could not be validated in its natural host, tional amyloids conferring bacterial colonies structural results obtained in vitro, in yeast and Escherichia coli integrity and cohesiveness, like curli [121], suggested clearly demonstrated its ability to populate amyloid- that prion-like conformational transitions to amyloid based prion conformations, becoming the first prion- structures might also exist in prokaryotes. like protein described in bacteria [30,60,63]. A bioinformatic survey of PrLD-containing proteins Soon after, the PrLD of the single-stranded DNA- in 839 different bacteria proteomes using PrionScan binding protein (SSB) from Campylobacter hominis identified 2200 prion-like candidates, accounting for a became the second experimentally validated bacterial 0.3% of the analyzed dataset [28]. Strikingly, this per- prion domain [125]. This PrLD can populate two alter- centage increases significantly when only pathogenic native conformations in E. coli, a soluble state without bacteria are considered. In Staphylococcus aureus, a prion activity and an amyloid-like prion conformation virulent pathogen that is currently the most common (SSB cPrD) that can be propagated over at least 100 cause of infections in hospitalized patients, predicted generations. prion-like proteins are 18% of the proteome. This sug- When the evolutionary behavior of PrLD’s in bacte- gests a role of PrLDs in pathogenesis, something con- ria is analyzed in detail, a pattern of apparent sporadic sistent with the fact that many microbial functional conservation is observed, which is often coupled to a amyloids are utilized by pathogens for invasion and wide distribution across multiple phyla [126], with Rho maintenance of infection. and SSB prion-like families displaying both features. Regarding the biological processes, a Gene Ontology This suggests that these domains are not necessarily (GO) terms analysis of these PrLD-containing proteins needed to develop the primary function of the protein. revealed an enrichment in cell morphogenesis (cell pro- Their presence may be occasionally beneficial, but jection and cell wall dynamics), in secretion, in inva- eventually, they may become detrimental to the fitness sion and virulence, and in nucleotide metabolism. A of a given species and subsequently purged out. Alter- survey of the Pfam families in such prion-like proteins natively, it could be that we are not detecting these revealed an enrichment in nucleotide binding domains; domains simply because they present compositions however, they differ from those found in eukaryotes that differ from those of the prion-like domains used and include those in GTP-binding elongation factors to train the algorithms. In this context, it should be or the Rho termination factors. Other enriched Pfam noted that the algorithms used for the identification of domains are connected to cell wall dynamics, cell wall prions in bacteria are trained on top of yeast prions metabolism, attachment to the cell wall, and secretion sequences, while it has been described that bacteria and invasion processes. Therefore, as it occurs in generally have a lower proportion of Q/N-enriched eukaryotes, bacterial PrLDs seem to be constituents of regions [25]. proteins involved in the regulation of gene expression Overall, the studies detailed in this section provide and transcription. The downstream processes they con- support for the hypothesis that the presence of prion- trol differ significantly, playing essential roles in speci- like proteins in the bacterial domain of life is not anec- fic functions at the periphery of the cell, such as the dotic, and they are likely a source of phenotypic diver- establishment of cell contacts to form multicellular sity that allows adaptation to defying environments. associations and biofilms [122,123], which protect However, many questions are still open, and much of against stress and antibacterial agents, while facilitat- bacteria’s prionic landscape remains to be understood. ing the attachment to host cells, explaining why prion- like proteins seem to be more prevalent in pathogens. The viral proteomes With the above data in hand, Pallares et al. [60] combined composition-based and soft amyloid If eukaryotic and bacterial cells bear prion-like pro- analysis-based approaches to scrutiny the prionic con- teins, why not the viruses that infect them? This was tent of the Gram-positive, anaerobic pathogen the question Tetz and Tetz [34,35] addressed by ana- Clostridium botulinum pathogen, identifying 54 pro- lyzing the prionic-load of bacteriophages and eukary- teins that accounted for a 1.5% of the whole pro- otic viruses using PLAAC. They detected > 5000 and teome. As in the previous study, one of the top ranked > 2500 putative prion-like polypeptides in bacterio- candidates was the Rho termination factor, which phages and eukaryotic viruses sequenced genomes, indeed contains a PrLD in > 20 different bacterial spe- respectively. In eukaryotic viruses prion-like candi- cies. The Rho factor is a conserved hexameric helicase dates were more prevalent in DNA than in RNA 2410 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes viruses and in enveloped viruses relative to nonen- in the SARS-CoV-2 spike protein might influence the veloped ones. affinity of this viral protein for angiotensin-converting Functional annotation analysis evidenced an enrich- enzyme 2 in human host cells [129]. ment in virus-specific processes involved in the estab- lishment of interactions with the host cell. In the case Conclusions of bacteriophages, the predominant functions corre- sponded to the attachment and penetration inside the The proteome studies discussed in this review make bacteria. It remains unknown if these processes are clear that the list of identified proteins bearing PrLDs also dependent on the presence of the peripheral will be steadily increasing and that they are widespread prion-like proteins predicted in bacteria. If this is the in all life forms, even in viruses. As a trend, PrLDs case, it will constitute a unique case of PrLDs medi- function as gene expression regulatory devices; how- ated cross-talk between different kingdoms of life. Not ever, these predictions have unveiled unprecedented surprisingly, prion-like proteins in eukaryotic viruses and fascinating examples of prion-mediated physiologi- were related to the adhesion and entry of viral genetic cal functions. PrLDs emerge as proteinaceous entities material inside the host. Of course, a large fraction of that encode molecular information essential to ensure PrLD-containing proteins was involved in the binding complex cellular events such as cell cycle regulation, and replication of nucleic acids, which against stems immune response, inflammation, long-term memory, from the necessary association between these disor- circadian clock regulation, host adaptation, cell adapt- dered regions and DNA/RNA binding domains. ability, and invasion, these latter functions being more Overall, PrLD-mediated switching on/off of regula- prevalent in pathogens. tory proteins seems to be a generic and conserved Proteins hosting PrLDs are modular and multifunc- mechanism, whereas the activated/inactivated process tional polypeptides that establish both homotypic and or function would be specific for any particular kind heterotypic protein–protein interactions, preferentially of organism. Consistent with this view, Nan and co- with other PrLDs. It is likely a synergy between these workers reported the first viral protein that behaves as features that links PrLDs to their biological functions. a prion [127]. The AcMNPV virus-encoded LEF-10 The transient and weak noncovalent interactions that protein works as a late expression factor essential for allow PrLDs to sample different conformational states tDNA replication and gene expression [128]. Its prion appear to overlap somehow with those responsible for behavior has been corroborated in yeast and in the the liquid liquid phase separation of different natural host, insect cells. Inside its native host, LEF-10 biomolecular condensates. Mutations in these regions can populate two different conformations, a soluble that impact these contacts’ strength/number can trigger and active one, and an aggregated and inactive state a transition from highly dynamic liquid condensates to that results in a reduction of late gene expression. This more solid-like aggregates, abrogating their dynamic/ conformational conversion is highly dependent of the promiscuous nature, leading to the onset of devastat- number of virions in the host cell, in such a way that ing protein aggregation diseases. From an evolutionary LEF-10 acts as a self-limiting factor becoming inactive point of view, the presence of these inherently risky when the presence of new virions is sufficient. It is domains in multiple species across kingdoms is likely worth to mention that the LEF-10 prion domain does the result of a positive natural selection. Bioinformat- not conform to the conventional conformational bias, ics approaches have been invaluable in their identifica- escaping from the identification of the above-described tion and characterization. Nonetheless, prion-like predictors. This is not an exception since several well- proteins that cannot be identified by compositional- characterized prions lack these domains [9], which and/or soft amyloid-based PrLD predictors have been argues for the development of more accurate algo- described [104-106]. These nonamyloidogenic proteins rithms since otherwise a significant fraction of the illustrate the co-existence of different flavors of prions prion sequential space would remain obscure. and challenge our knowledge of the compositional/se- Viruses behave as obligate parasites, in such a way quential determinants of beneficial and pathological their survival and replication in host cells requires prion-like transitions. An overlooked possibility is that adaptation to cellular conditions. The use of prion the sequence features enabling prion-like behavior mechanisms for a fast regulation of gene expression could exhibit a significant degree of species-specificity. might help viruses to quickly respond to the response In this scenario, a low Q/N-content in any given spe- to their presence elicited by host. Hence, we anticipate cies would be immediately translated into a decreased that more viral prion-like proteins will be discovered PrLD predicted content [25], independently if this true soon and indeed a recent study suggests that a PrLD or not. Further experimental characterizations in novel FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2411 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. organisms are needed to fuel novel prion-like predic- cell survival under environmental stress. Science 336, tive strategies that flee from our present yeast-centric 355–359. view. 10 March ZM, King OD and Shorter J (2016) Prion-like This is important not only because they will allow domains as epigenetic regulators, scaffolds for us to deep into the biology of the prion-like proteins subcellular organization, and drivers of but also for their potential therapeutical and nanotech- neurodegenerative disease. Brain Res 1647, 9–18. nological implications. In this last respect, the knowl- 11 Osherovich LZ and Weissman JS (2001) Multiple Gln/ edge we have gained from proteome-wide analyses has Asn-rich prion domains confer susceptibility to induction of the yeast [PSI(+)] prion. Cell 106, 183– already crystallized in the development of new nano- 194. materials with fascinating functional properties 12 Ross ED, Edskes HK, Terry MJ and Wickner RB [130,131] that would have never seen the light without (2005) Primary sequence independence for prion the herein described bioinformatic analysis. formation. Proc Natl Acad Sci USA 102, 12825– 12830. Acknowledgements 13 Alberti S, Halfmann R, King O, Kapila A and Lindquist S (2009) A systematic survey identifies This work was funded by the Spanish Ministry of prions and illuminates sequence features of Science and Innovation PID2019-105017RB-100 to SV prionogenic proteins. Cell 137, 146–158. and by ICREA, ICREA-Academia 2020, to SV. MG- 14 Toombs JA, McCarty BR and Ross ED (2010) G was supported by the Spanish Ministry of Science Compositional determinants of prion formation in and Innovation via a doctoral grant (FPU16/02465). yeast. Mol Cell Biol 30, 319–332. Graphical abstract and Fig. 2 were created with BioRe 15 Cox BS (1965) Ψ, A cytoplasmic suppressor of super- nder.com. suppressor in yeast. Heredity 20, 505–521. 16 Aigle M and Lacroute F (1975) Genetical aspects of Conflict of interest [URE3], a non-mitochondrial, cytoplasmically inherited mutation in yeast. Mol Gen Genet 136, 327– The authors declare no conflict of interest. 335. 17 Masison DC and Wickner RB (1995) Prion-inducing domain of yeast Ure2p and protease resistance of References Ure2p in prion-containing cells. Science 270, 93–95. 1 Sikorska B and Liberski PP (2012) Human prion 18 Tuite MF, Staniforth GL and Cox BS (2015) [PSI(+)] diseases: from Kuru to variant Creutzfeldt-Jakob turns 50. Prion 9, 318–332. disease. Subcell Biochem 65, 457–496. 19 Zambrano R, Conchillo-Sole O, Iglesias V, Illa R, 2 Si K (2015) Prions: what are they good for? Annu Rev Rousseau F, Schymkowitz J, Sabate R, Daura X and Cell Dev Biol 31, 149–169. Ventura S (2015) PrionW: a server to identify proteins 3 Shorter J and Lindquist S (2005) Prions as adaptive containing glutamine/asparagine rich prion-like conduits of memory and inheritance. Nat Rev Genet 6, domains and their amyloid cores. Nucleic Acids Res 435–450. 43, W331–W337. 4 Aguzzi A and Calella AM (2009) Prions: protein 20 Afsar Minhas FUA, Ross ED and Ben-Hur A (2017) aggregation and infectious diseases. Physiol Rev 89, Amino acid composition predicts prion activity. PLoS 1105–1152. Comput Biol 13, e1005465. 5 Ziaunys M, Sneideris T and Smirnovas V (2020) 21 Espinosa Angarica V, Ventura S and Sancho J (2013) Formation of distinct prion protein amyloid fibrils Discovering putative prion sequences in complete under identical experimental conditions. Sci Rep 10, proteomes using probabilistic representations of Q/N- 4572. rich domains. BMC Genom 14, 316. 6 Holec SAM and Woerman AL (2021) Evidence of 22 Espinosa Angarica V, Angulo A, Giner A, Losilla G, distinct a-synuclein strains underlying disease Ventura S and Sancho J (2014) PrionScan: an online heterogeneity. Acta Neuropathol 142, 73–86. database of predicted prion domains in complete 7 Chien P, Weissman JS and DePace AH (2004) proteomes. BMC Genom 15, 102. Emerging principles of conformation-based prion 23 Harrison PM and Gerstein M (2003) A method to inheritance. Annu Rev Biochem 73, 617–656. assess compositional bias in biological sequences and 8 Liebman SW and Chernoff YO (2012) Prions in yeast. its application to prion-like glutamine/asparagine-rich Genetics 191, 1041–1072. domains in eukaryotic proteomes. Genome Biol 4, R40. 9 Suzuki G, Shimazu N and Tanaka M (2012) A yeast 24 Lancaster AK, Nutter-Upham A, Lindquist S and prion, Mod5, promotes acquired drug resistance and King OD (2014) PLAAC: a web and command-line 2412 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes application to identify proteins with prion-like amino 40 Gotor NL, Armaos A, Calloni G, Torrent Burgas M, acid composition. Bioinformatics 30, 2501–2502. Vabulas RM, De Groot NS and Tartaglia GG (2020) 25 Michelitsch MD and Weissman JS (2000) A census of RNA-binding and prion domains: the Yin and Yang glutamine/asparagine-rich regions: implications for of phase separation. Nucleic Acids Res 48, 9491–9504. their conserved function and the prediction of novel 41 Martin EW, Holehouse AS, Peran I, Farag M, Incicco prions. Proc Natl Acad Sci USA 97, 11910–11915. JJ, Bremer A, Grace CR, Soranno A, Pappu RV and 26 Sabate R, Rousseau F, Schymkowitz J and Ventura S Mittag T (2020) Valence and patterning of aromatic (2015) What makes a protein sequence a prion? PLoS residues determine the phase behavior of prion-like Comput Biol 11, e1004013. domains. Science 367, 694–699. 27 Toombs JA, Petri M, Paul KR, Kan GY, Ben-Hur A 42 Monahan Z, Ryan VH, Janke AM, Burke KA, and Ross ED (2012) De novo design of synthetic prion Rhoads SN, Zerze GH, O’Meally R, Dignon GL, domains. Proc Natl Acad Sci USA 109, 6519–6524. Conicella AE, Zheng W et al. (2017) Phosphorylation 28 Iglesias V, de Groot NS and Ventura S (2015) of the FUS low-complexity domain disrupts phase Computational analysis of candidate prion-like separation, aggregation, and toxicity. EMBO J 36, proteins in bacteria and their role. Front Microbiol 6, 2951–2967. 1123. 43 Patel A, Lee HO, Jawerth L, Maharana S, Jahnel M, 29 Chakrabortee S, Kayatekin C, Newby GA, Mendillo Hein MY, Stoynov S, Mahamid J, Saha S, Franzmann ML, Lancaster A and Lindquist S (2016) TM et al. (2015) A liquid-to-solid phase transition of Luminidependens (LD) is an Arabidopsis protein with the ALS protein FUS accelerated by disease mutation. prion behavior. Proc Natl Acad Sci USA 113, 6065– Cell 162, 1066–1077. 6070. 44 Prasad A, Bharathi V, Sivalingam V, Girdhar A and 30 Yuan AH and Hochschild A (2017) A bacterial global Patel BK (2019) Molecular mechanisms of TDP-43 regulator forms a prion. Science 355, 198–201. misfolding and pathology in amyotrophic lateral 31 An L and Harrison PM (2016) The evolutionary scope sclerosis. Front Mol Neurosci 12, 25. and neurological disease linkage of yeast-prion-like 45 Kim HJ, Kim NC, Wang Y-D, Scarborough EA, proteins in humans. Biol Direct 11, 32. Moore J, Diaz Z, MacLea KS, Freibaum B, Li S, 32 Batlle C, de Groot NS, Iglesias V, Navarro S and Molliex A et al. (2013) Mutations in prion-like domains Ventura S (2017) Characterization of soft amyloid in hnRNPA2B1 and hnRNPA1 cause multisystem cores in human prion-like proteins. Sci Rep 7, 12134. proteinopathy and ALS. Nature 495, 467–473. 33 Pallares I, de Groot NS, Iglesias V, Sant’Anna R, 46 Fernandez MR, Batlle C, Gil-Garcıa M and Ventura S Biosca A, Fern andez-Busquets X and Ventura S (2018) (2017) Amyloid cores in prion domains: key regulators Discovering putative prion-like proteins in Plasmodium for prion conformational conversion. Prion 11, 31–39. falciparum: a computational and experimental analysis. 47 Linding R, Schymkowitz J, Rousseau F, Diella F and Front Microbiol 9, 1737. Serrano L (2004) A comparative study of the 34 Tetz G and Tetz V (2017) Prion-like domains in relationship between protein structure and beta- phagobiota. Front Microbiol 8, 2239. aggregation in globular and intrinsically disordered 35 Tetz G and Tetz V (2018) Prion-like domains in proteins. J Mol Biol 342, 345–353. eukaryotic viruses. Sci Rep 8, 8931. 48 Sabate R, Rousseau F, Schymkowitz J, Batlle C and 36 Iglesias V, Paladin L, Juan-Blanco T, Pallares I, Aloy Ventura S (2015) Amyloids or prions? That is the P, Tosatto SCE and Ventura S (2019) In silico question. Prion 9, 200–206. characterization of human prion-like proteins: beyond 49 DePace AH, Santoso A, Hillner P and Weissman JS neurological diseases. Front Physiol 10, 314. (1998) A critical role for amino-terminal glutamine/ 37 King OD, Gitler AD and Shorter J (2012) The tip of asparagine repeats in the formation and propagation the iceberg: RNA-binding proteins with prion-like of a yeast prion. Cell 93, 1241–1252. domains in neurodegenerative disease. Brain Res 1462, 50 Perutz MF, Johnson T, Suzuki M and Finch JT (1994) 61–80. Glutamine repeats as polar zippers: their possible role 38 Malinovska L, Kroschwald S and Alberti S (2013) in inherited neurodegenerative diseases. Proc Natl Protein disorder, prion propensities, and self- Acad Sci USA 91, 5355–5358. organizing macromolecular collectives. Biochim 51 Prilusky J, Felder CE, Zeev-Ben-Mordehai T, Rydberg Biophys Acta 1834, 918–931. EH, Man O, Beckmann JS, Silman I and Sussman JL 39 Franzmann TM, Jahnel M, Pozniakovsky A, (2005) FoldIndex: a simple tool to predict whether a Mahamid J, Holehouse AS, N€ uske E, Richter D, given protein sequence is intrinsically unfolded. Baumeister W, Grill SW, Pappu RV et al. (2018) Bioinformatics 21, 3435–3438. Phase separation of a yeast prion protein promotes 52 Cascarina SM and Ross ED (2020) Natural and cellular fitness. Science 359, eaao5654. pathogenic protein sequence variation affecting prion- FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2413 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. like domains within and across human proteomes. hnRNPA2 prion-like protein. FEBS Lett 591, 1966– BMC Genom 21, 23. 1971. 53 Batlle C, Yang P, Coughlin M, Messing J, 65 Iglesias V, Conchillo-Sole O, Batlle C and Ventura S Pesarrodona M, Szulc E, Salvatella X, Kim HJ, Taylor (2019) AMYCO: evaluation of mutational impact on JP and Ventura S (2020) hnRNPDL phase separation prion-like proteins aggregation propensity. BMC is regulated by alternative splicing and disease-causing Bioinformatics 20, 24. mutations accelerate its aggregation. Cell Rep 30, 66 Kim Y-C, Won S-Y and Jeong B-H (2020) 1117–1128.e5. Identification of prion disease-related somatic 54 Harbi D, Kumar M and Harrison PM (2011) LPS- mutations in the prion protein gene (PRNP) in cancer annotate: complete annotation of compositionally patients. Cells 9, 1480. biased regions in the protein knowledgebase. Database 67 Kim D-J, Kim Y-C, Kim A-D and Jeong B-H (2020) 2011, baq031. Novel polymorphisms and genetic characteristics of the 55 Harbi D, Parthiban M, Gendoo DM, Ehsani S, prion protein gene (PRNP) in dogs-a resistant animal Kumar M, Schmitt-Ulms G, Sowdhamini R and of prion disease. Int J Mol Sci 21, 4160. Harrison PM (2012) PrionHome: a database of prions 68 Kim Y-C, Won S-Y, Do K and Jeong B-H (2020) and other sequences relevant to prion phenomena. Identification of the novel polymorphisms and PLoS One 7, e31785. potential genetic features of the prion protein gene 56 Kryndushkin DS, Wickner RB and Tycko R (2011) (PRNP) in horses, a prion disease-resistant animal. Sci The core of Ure2p prion fibrils is formed by the Rep 10, 8926. N-terminal segment in a parallel cross-b structure: 69 Kim S-K, Kim Y-C, Won S-Y and Jeong B-H (2019) evidence from solid-state NMR. J Mol Biol 409, 263– Potential scrapie-associated polymorphisms of the 277. prion protein gene (PRNP) in Korean native black 57 Shewmaker F, Kryndushkin D, Chen B, Tycko R and goats. Sci Rep 9, 15293. Wickner RB (2009) Two prion variants of Sup35p 70 Batlle C, Iglesias V, Navarro S and Ventura S (2017) have in-register parallel b-sheet structures, independent Prion-like proteins and their computational of hydration. Biochemistry 48, 5074–5082. identification in proteomes. Expert Rev Proteomics 14, 58 Couthouis J, Hart MP, Erion R, King OD, Diaz Z, 335–350. Nakaya T, Ibrahim F, Kim H-J, Mojsilovic-Petrovic J, 71 Harvey ZH, Chen Y and Jarosz DF (2018) Protein- Panossian S et al. (2012) Evaluating the role of the based inheritance: epigenetics beyond the chromosome. FUS/TLS-related gene EWSR1 in amyotrophic lateral Mol Cell 69, 195–202. sclerosis. Hum Mol Genet 21, 2899–2911. 72 Levkovich SA, Rencus-Lazar S, Gazit E and Laor 59 Tariq M, Wegrzyn R, Anwar S, Bukau B and Paro R Bar-Yosef D (2021) Microbial prions: dawn of a new (2013) Drosophila GAGA factor polyglutamine domains era. Trends Biochem Sci 46, 391–405. exhibit prion-like behavior. BMC Genom 14, 374. 73 Cascarina SM and Ross ED (2014) Yeast prions and 60 Pallares I, Iglesias V and Ventura S (2015) The Rho human prion-like proteins: sequence features and termination factor of Clostridium botulinum contains a prediction methods. Cell Mol Life Sci 71, 2047–2063. prion-like domain with a highly amyloidogenic core. 74 Oamen HP, Lau Y and Caudron F (2020) Prion-like Front Microbiol 6, 1516. proteins as epigenetic devices of stress adaptation. Exp 61 Sant’Anna R, Fernandez MR, Batlle C, Navarro S, de Cell Res 396, 112262. Groot NS, Serpell L and Ventura S (2016) 75 Caudron F and Barral Y (2013) A super-assembly of Characterization of amyloid cores in prion domains. Whi3 encodes memory of deceptive encounters by Sci Rep 6, 34274. single cells during yeast courtship. Cell 155, 1244– 62 Duernberger Y, Liu S, Riemschoss K, Paulsen L, 1257. Bester R, Kuhn P-H, Sch€ olling M, Lichtenthaler SF 76 Couthouis J, Hart MP, Shorter J, DeJesus-Hernandez and Vorberg I (2018) Prion replication in the M, Erion R, Oristano R, Liu AX, Ramos D, Jethava mammalian cytosol: functional regions within a prion N, Hosangadi D et al. (2011) A yeast functional screen domain driving induction, propagation, and predicts new candidate ALS disease genes. Proc Natl inheritance. Mol Cell Biol 38, e00111-18. Acad Sci USA 108, 20881–20890. 63 Pallares I and Ventura S (2017) The transcription 77 Carlsten JOP, Zhu X and Gustafsson CM (2013) The terminator rho: a first bacterial prion. Trends multitalented mediator complex. Trends Biochem Sci Microbiol 25, 434–437. 38, 531–537. 64 Batlle C, Fernandez MR, Iglesias V and Ventura S 78 Shaikhibrahim Z, Offermann A, Halbach R, Vogel W, (2017) Perfecting prediction of mutational impact on Braun M, Kristiansen G, Bootz F, Wenzel J, Mikut R, the aggregation propensity of the ALS-associated Lengerke C et al. (2015) Clinical and molecular 2414 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes implications of MED15 in head and neck squamous prion-like RNA binding proteins. Cell 174, 688– cell carcinoma. Am J Pathol 185, 1114–1122. 699.e16. 79 Harrison AF and Shorter J (2017) RNA-binding 92 Verdile V, De Paola E and Paronetto MP (2019) proteins with prion-like domains in health and disease. Aberrant phase transitions: side effects and novel Biochem J 474, 1417–1438. therapeutic strategies in human disease. Front Genet 80 Louka A, Zacco E, Temussi PA, Tartaglia GG and 10, 173. Pastore A (2020) RNA as the stone guest of protein 93 Shorter J (2016) Engineering therapeutic protein aggregation. Nucleic Acids Res 48, 11880–11889. disaggregases. Mol Biol Cell 27, 1556–1560. 81 Neumann M, Sampathu DM, Kwong LK, Truax AC, 94 Gueroussov S, Weatheritt RJ, O’Hanlon D, Lin Z-Y, Micsenyi MC, Chou TT, Bruce J, Schuck T, Narula A, Gingras A-C and Blencowe BJ (2017) Grossman M, Clark CM et al. (2006) Ubiquitinated Regulatory expansion in mammals of multivalent TDP-43 in frontotemporal lobar degeneration and hnRNP assemblies that globally control alternative amyotrophic lateral sclerosis. Science 314, 130–133. splicing. Cell 170, 324–339.e23. 82 Da Cruz S and Cleveland DW (2011) Understanding 95 Deshaies J-E, Shkreta L, Moszczynski AJ, Sidibe H, the role of TDP-43 and FUS/TLS in ALS and beyond. Semmler S, Fouillen A, Bennett ER, Bekenstein U, Curr Opin Neurobiol 21, 904–919. Destroismaisons L, Toutant J et al. (2018) TDP-43 83 Sun Z, Diaz Z, Fang X, Hart MP, Chesi A, Shorter J regulates the alternative splicing of hnRNP A1 to yield and Gitler AD (2011) Molecular determinants and an aggregation-prone variant in amyotrophic lateral genetic modifiers of aggregation and toxicity for the sclerosis. Brain 141, 1320–1333. ALS disease protein FUS/TLS. PLoS Biol 9, e1000614. 96 Gill J, Park Y, McGinnis JP, Perez-Sanchez C, 84 Chen C, Ding X, Akram N, Xue S and Luo S-Z Blanchette M and Si K (2017) Regulated intron (2019) Fused in sarcoma: properties, self-assembly and removal integrates motivational state and experience. correlation with neurodegenerative diseases. Molecules Cell 169, 836–848.e15. 24, 1622. 97 Hofweber M and Dormann D (2019) Friend or foe- 85 Ticozzi N, Vance C, Leclerc AL, Keagle P, Glass JD, post-translational modifications as regulators of phase McKenna-Yasek D, Sapp PC, Silani V, Bosco DA, separation and RNP granule dynamics. J Biol Chem Shaw CE et al. (2011) Mutational analysis reveals the 294, 7137–7150. FUS homolog TAF15 as a candidate gene for familial 98 Owen I and Shewmaker F (2019) The role of post- amyotrophic lateral sclerosis. Am J Med Genet B translational modifications in the phase transitions of Neuropsychiatr Genet 156B, 285–290. intrinsically disordered proteins. Int J Mol Sci 20, 86 Shin Y and Brangwynne CP (2017) Liquid phase 5501. condensation in cell physiology and disease. Science 99 Dormann D, Madl T, Valori CF, Bentmann E, 357, eaaf4382. Tahirovic S, Abou-Ajram C, Kremmer E, Ansorge O, 87 Boeynaems S, Alberti S, Fawzi NL, Mittag T, Mackenzie IRA, Neumann M et al. (2012) Arginine Polymenidou M, Rousseau F, Schymkowitz J, Shorter methylation next to the PY-NLS modulates J, Wolozin B, Van Den Bosch L et al. (2018) Protein transportin binding and nuclear import of FUS. phase separation: a new phase in cell biology. Trends EMBO J 31, 4258–4275. Cell Biol 28, 420–435. 100 Suarez-Calvet M, Neumann M, Arzberger T, Abou- 88 Gomes E and Shorter J (2019) The molecular language Ajram C, Funk E, Hartmann H, Edbauer D, Kremmer of membraneless organelles. J Biol Chem 294, 7115– E, G€ obl C, Resch M et al. (2016) Monomethylated and 7127. unmethylated FUS exhibit increased binding to 89 Molliex A, Temirov J, Lee J, Coughlin M, Kanagaraj transportin and distinguish FTLD-FUS from ALS- AP, Kim HJ, Mittag T and Taylor JP (2015) Phase FUS. Acta Neuropathol 131, 587–604. separation by low complexity domains promotes stress 101 Antonets KS, Belousov MV, Sulatskaya AI, granule assembly and drives pathological fibrillization. Belousova ME, Kosolapova AO, Sulatsky MI, Cell 163, 123–133. Andreeva EA, Zykin PA, Malovichko YV, Shtark OY 90 Lin Y, Protter DSW, Rosen MK and Parker R (2015) et al. (2020) Accumulation of storage proteins in plant Formation and maturation of phase-separated liquid seeds is mediated by amyloid formation. PLoS Biol 18, droplets by RNA-binding proteins. Mol Cell 60, 208– e3000564. 219. 102 Santos J and Ventura S (2021) Functional amyloids 91 Wang J, Choi J-M, Holehouse AS, Lee HO, Zhang X, germinate in plants. Trends Plant Sci 26, 7–10. Jahnel M, Maharana S, Lemaitre R, Pozniakovsky A, 103 Brown JCS and Lindquist S (2009) A heritable switch Drechsel D et al. (2018) A molecular grammar in carbon source utilization driven by an unusual yeast governing the driving forces for phase separation of prion. Genes Dev 23, 2320–2332. FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2415 Federation of European Biochemical Societies Prediction of prion-like proteins in proteomes M. Gil-Garcia et al. 104 Chakrabortee S, Byers JS, Jones S, Garcia DM, secretome in malarial infection. Science 306, 1934– Bhullar B, Chang A, She R, Lee L, Fremin B, 1937. Lindquist S et al. (2016) Intrinsically disordered 117 Balaji S, Babu MM, Iyer LM and Aravind L (2005) proteins drive emergence and inheritance of biological Discovery of the principal specific transcription factors traits. Cell 167, 369–381.e12. of Apicomplexa and their implication for the evolution 105 Harvey ZH, Chakravarty AK, Futia RA and Jarosz of the AP2-integrase DNA binding domains. Nucleic DF (2020) A prion epigenetic switch establishes an Acids Res 33, 3994–4006. active chromatin state. Cell 180, 928–940.e14. 118 Singh S, Santos JM, Orchard LM, Yamada N, van 106 Chakravarty AK, Smejkal T, Itakura AK, Garcia DM Biljon R, Painter HJ, Mahony S and Llinas M (2020) and Jarosz DF (2020) A non-amyloid prion particle The PfAP2-G2 transcription factor is a critical that activates a heritable gene expression program. regulator of gametocyte maturation. Mol Microbiol Mol Cell 77, 251–265.e9. 115, 1005–1024. 107 Lee I, Aukerman MJ, Gore SL, Lohman KN, 119 Toro-Moreno M, Sylvester K, Srivastava T, Posfai D Michaels SD, Weaver LM, John MC, Feldmann KA and Derbyshire ER (2020) RNA-seq analysis and Amasino RM (1994) Isolation of illuminates the early stages of Plasmodium liver LUMINIDEPENDENS: a gene involved in the infection. MBio 11, e03234-19. control of flowering time in Arabidopsis. Plant Cell 6, 120 Akkaya M, Bansal A, Sheehan PW, Pena M, Molina- 75–83. Cruz A, Orchard LM, Cimperman CK, Qi C-F, Ross 108 Huang H and Nusinow DA (2016) Into the evening: P, Yazew T et al. (2020) A single-nucleotide complex interactions in the Arabidopsis circadian polymorphism in a Plasmodium berghei ApiAP2 clock. Trends Genet 32, 674–686. transcription factor alters the development of host 109 Jung J-H, Barbosa AD, Hutin S, Kumita JR, Gao M, immunity. Sci Adv 6, eaaw6957. Derwort D, Silva CS, Lai X, Pierre E, Geng F et al. 121 Chapman MR, Robinson LS, Pinkner JS, Roth R, (2020) A prion-like domain in ELF3 functions as a Heuser J, Hammar M, Normark S and Hultgren SJ thermosensor in Arabidopsis. Nature 585, 256–260. (2002) Role of Escherichia coli curli operons in 110 Wilkinson EG and Strader LC (2020) A prion-based directing amyloid fiber formation. Science 295, 851– thermosensor in plants. Mol Cell 80, 181–182. 855. 111 Snow RW, Guerra CA, Noor AM, Myint HY and 122 de la Fuente-N un~ez C, Reffuveille F, Fernandez L and Hay SI (2005) The global distribution of clinical Hancock REW (2013) Bacterial biofilm development episodes of Plasmodium falciparum malaria. Nature as a multicellular adaptation: antibiotic resistance and 434, 214–217. new therapeutic strategies. Curr Opin Microbiol 16, 112 Singh GP, Chandra BR, Bhattacharya A, Akhouri 580–589. RR, Singh SK and Sharma A (2004) Hyper-expansion 123 Sharma D, Misba L and Khan AU (2019) of asparagines correlates with an abundance of Antibiotics versus biofilm: an emerging battleground proteins with prion-like domains in Plasmodium in microbial communities. Antimicrob Resist Infect falciparum. Mol Biochem Parasitol 137, 307–319. Control 8, 76. 113 Muralidharan V, Oksman A, Pal P, Lindquist S and 124 Skordalakes E and Berger JM (2003) Structure of the Goldberg DE (2012) Plasmodium falciparum heat Rho transcription terminator: mechanism of mRNA shock protein 110 stabilizes the asparagine repeat-rich recognition and helicase loading. Cell 114, 135–146. parasite proteome during malarial fevers. Nat Commun 125 Fleming E, Yuan AH, Heller DM and Hochschild A 3, 1310. (2019) A bacteria-based genetic assay detects prion 114 Malinovska L, Palm S, Gibson K, Verbavatz J-M and formation. Proc Natl Acad Sci USA 116, 4605–4610. Alberti S (2015) Dictyostelium discoideum has a highly 126 Harrison PM (2019) Evolutionary behaviour of Q/N-rich proteome and shows an unusual resilience to bacterial prion-like proteins. PLoS One 14, e0213030. protein aggregation. Proc Natl Acad Sci USA 112, 127 Nan H, Chen H, Tuite MF and Xu X (2019) A viral E2620–E2629. expression factor behaves as a prion. Nat Commun 10, 115 Santarriaga S, Haver HN, Kanack AJ, Fikejs AS, 359. Sison SL, Egner JM, Bostrom JR, Seminary ER, Hill 128 Lu A and Miller LK (1995) The roles of eighteen RB, Link BA et al. (2018) SRCP1 conveys resistance baculovirus late expression factor genes in to polyglutamine aggregation. Mol Cell 71, 216– transcription and DNA replication. J Virol 69, 975– 228.e7. 982. 116 Hiller NL, Bhattacharjee S, van Ooij C, Liolios K, 129 Tetz G and Tetz V (2020) SARS-CoV-2 prion-like Harrison T, Lopez-Estra~ no C and Haldar K (2004) A domains in spike proteins enable higher affinity to host-targeting signal in virulence proteins reveals a ACE2 [PREPRINT]. 2416 FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies M. Gil-Garcia et al. Prediction of prion-like proteins in proteomes 130 Dıaz-Caballero M, Navarro S, Fuentes I, Teixidor F 136 Harrison PM (2017) fLPS: Fast discovery of and Ventura S (2018) Minimalist prion-inspired polar compositional biases for the protein universe. BMC self-assembling peptides. ACS Nano 12, 5394–5407. Bioinformatics 18, 476. 131 Dıaz-Caballero M, Navarro S, Nuez-Martınez M, 137 Maurer-Stroh S, Debulpaep M, Kuemmerer N, Lopez Peccati F, Rodrıguez-Santiago L, Sodupe M, Teixidor de la Paz M, Martins IC, Reumers J, Morris KL, F and Ventura S (2021) pH-responsive self-assembly Copland A, Serpell L, Serrano L et al. (2010) of amyloid fibrils for dual hydrolase-oxidase reactions. Exploring the sequence determinants of amyloid ACS Catalysis 11, 595–607. structure using position-specific scoring matrices. Nat 132 Conchillo-Sole O, de Groot NS, Aviles FX, Vendrell Methods 7, 237–242. J, Daura X and Ventura S (2007) AGGRESCAN: a 138 Gilks N, Kedersha N, Ayodele M, Shen L, Stoecklin server for the prediction and evaluation of “hot spots” G, Dember LM and Anderson P (2004) Stress granule of aggregation in polypeptides. BMC Bioinformatics 8, assembly is mediated by prion-like aggregation of 65. TIA-1. Mol Biol Cell 15, 5383–5398. 133 Wojciechowski JW and Kotulska M (2020) PATH – 139 Paushkin SV, Kushnirov V, Smirnov VN and Ter- prediction of amyloidogenicity by threading and Avanesyan MD (1997) In vitro propagation of the machine learning. Sci Rep 10, 7721. prion-like state of yeast Sup35 protein. Science 277, 134 Niu M, Li Y, Wang C and Han K (2018) RFAmyloid: 381–383. a web server for predicting amyloid proteins. Int J 140 Wickner RB (1994) [URE3] as an altered URE2 Mol Sci 19, 2071. protein: evidence for a prion analog in Saccharomyces 135 Burdukiewicz M, Sobczyk P, R€ odiger S, Duda-Madej cerevisiae. Science 264, 566–569. A, Mackiewicz P and Kotulska M (2017) 141 Sondheimer N and Lindquist S (2000) Rnq1: an Amyloidogenic motifs revealed by n-gram analysis. Sci epigenetic modifier of protein function in yeast. Mol Rep 7, 12961. Cell 5, 163–172. FEBS Open Bio 11 (2021) 2400–2417 ª 2021 The Authors. FEBS Open Bio published by John Wiley & Sons Ltd on behalf of 2417 Federation of European Biochemical Societies