A viral expression factor behaves as a prion

Prions are proteins that can fold into multiple conformations some of which are self-propagating. Such prion-forming proteins have been found in animal, plant, fungal and bacterial species, but have not yet been identified in viruses. Here we report that LEF-10, a baculovirus-encoded protein, behaves as a prion. Full-length LEF-10 or its candidate prion-forming domain (cPrD) can functionally replace the PrD of Sup35, a widely studied prion-forming protein from yeast, displaying a [PSI+]-like phenotype. Furthermore, we observe that high multiplicity of infection can induce the conversion of LEF-10 into an aggregated state in virus-infected cells, resulting in the inhibition of viral late gene expression. Our findings extend the knowledge of current prion proteins from cellular organisms to non-cellular life forms and provide evidence to support the hypothesis that prion-forming proteins are a widespread phenomenon in nature. Prion-forming proteins have been found in animals, plants, fungi, and bacteria. Here, Nan et al. report that a baculovirus-encoded protein behaves as prion in a yeast system and forms aggregates at high multiplicity of infection in insect cells that affect baculovirus replication.

P rions are proteins that can take up multiple conformations that show distinct aggregation propensities and where at least one of the conformations is self-propagating. In mammals, the prion protein PrP can induce the formation of amyloid plaques in the central nervous system and cause neurodegenerative diseases 1,2 . Such prion diseases can affect diverse species, e.g., kuru, Creutzfeldt-Jacob disease, fatal familial insomnia, and Gerstmann-Sträussler-Scheinker disease in humans, bovine spongiform encephalopathy in cattle, scrapie in sheep and goats, and chronic wasting disease in deer and elk. Sharing common characteristics with PrP, a number of prionforming proteins have been identified in yeast and filamentous fungi: [ 8 , and [Het-s] 9 . Fungal prions are not typically detrimental to their hosts 10 , but they are useful in helping us to understand how mammalian prions are propagated. Recently, prion-forming proteins were discovered in plant 11 and bacteria 12,13 . Most non-mammalian prions, if not all, serve as modulators of non-pathogenic, epigenetic phenotypes and in several cases they may actually play important roles in normal biological processes, providing growth advantages under certain conditions [14][15][16] . For example, Luminidependens (LD), the first identified plant protein behaving as a prion, is a transcriptional regulator involved in the vernalization 11,17 . Such a selfpropagating conformational change in LD is considered to be a signal that can switch plants between reproductive and vegetative growth 11 . This suggests that prion-like conversion can act as a sensor, responding to environmental change, and subsequently regulate gene expression and cellular processes.
It has long been recognized that prion-based mechanisms provide a plausible regulatory pathway for a host's response to environmental stress. The rate of induction of the [PSI + ] prion in yeast, where it acts as an epigenetic modifier of translation termination, increases when cells are exposed to oxidative stress or high salt conditions 18 . The yeast prion [MOT3 + ], which regulates the expression of cell wall biogenesis-related genes, enables the cell to tolerate high concentrations of ethanol 19 . [MOD + ], a prion state of the yeast tRNA isopentenyl transferase, contributes to the resistance of yeast cell to antifungal drugs such as fluconazole and ketoconazole 8 . Also, prion-like conformation of mitochondrial antiviral signaling protein (MAVS) and the apoptosis-associated protein ASC in mammalian cells activate the downstream immunological or inflammatory signaling pathways 20,21 . These findings suggest that the ability to switch a protein to a heritable prion form under certain conditions may offer the host evolutionary advantages.
The potential evolutionary advantage suggests that prion proteins should exist widely in various types of organisms. Viruses are the most abundant form of life on earth. However, a virus-encoded prion has not been experimentally identified, although prion-like domains in phage have been predicted by bioinformatic methods 22 . Autographa californica multiple nucleopolyhedrovirus (AcMNPV) is a large double-stranded DNA virus that infects insects. It is the archetype species of the genus Alphabaculovirus of the family Baculoviridae, and is widely used as a vector for the production of recombinant proteins in the pharmaceutical industry. The AcMNPVencoded LEF-10 protein has been identified as a viral late expression factor 23 and is thought to be one of the genes positively selected for during baculovirus evolution 24 . Although LEF-10 in Bombyx mori nucleopolyhedrovirus (BmNPV) and AcMNPV have been identified as being essential for the viral DNA replication and gene expression 25,26 , the exact role of LEF-10 in virus infection has not been extensively studied. We have previously reported that LEF-10 can form unique detergent (SDS)-resistant, high molecular weight aggregates 27 and that the aggregation of LEF-10 can be induced by overproduction leading to subsequent down-regulation of viral late gene expression 25 . These characteristics of the LEF-10 protein mirror those of a prion protein. In this study, we have investigated the prion behavior of LEF-10 in a heterologous (yeast) system and in virus-infected insect cells.

Results
Inducible aggregation of LEF-10 in infected cells. To investigate the aggregation behavior of LEF-10 under physiological conditions, we developed a recombinant baculovirus sensor system transmitting the aggregation state of LEF-10 to a downstream gene output. The sensor system comprises a LEF-10-EGFP driven by the native lef-10 promoter and an mCherry reporter regulated by a baculovirus very late promoter (the p10 promoter) ( Fig. 1a and Supplementary Fig. 1e). In living baculovirus-infected Spodoptera frugiperda (Sf9) cells, LEF-10-EGFP shows two distinct distribution patterns, diffused or aggregated. The expression levels of mCherry were high in cells with evenly diffused LEF-10-EGFP, but the fluorescence of mCherry was undetectable in the cells containing LEF-10-EGFP aggregates 25 (Fig. 1b). Further computational analysis showed that once such aggregation occurred, most of the LEF-10 protein concentrated in small areas in the infected cells (Fig. 1c). In our designed genetic sensor system, correlation of the LEF-10-EGFP state with its regulated mCherry readout implied that aggregation of LEF-10 limited the availability of the fusion protein to activate the AcMNPV late promoter thereby reducing the expression of the mCherry reporter. We speculate that aggregated nuclei of LEF-10-EGFP could be inducing soluble LEF-10-EGFP to adopt the aggregated form, resulting in depletion of soluble, functional LEF-10-EGFP and the subsequent repression of late gene expression. The aggregation propensity of LEF-10-EGFP in infected cells is consistent with the behavior described for previously identified prions 28,29 .
LEF-10 behaves as a prion in a yeast prion reporter assay. In order to verify the prion characteristics of LEF-10, we employed an assay based on the well-characterized prion phenotypes of the Saccharomyces cerevisiae translation termination factor Sup35 7 . This protein consists of an N-terminal modular prion-forming domain (PrD), a highly charged middle region (M) and a Cterminal release function domain (C). To determine whether LEF-10 could produce an epigenetic modification of the heterologous functional protein, we substituted the PrD of Sup35 with LEF-10 to generate a LEF-10-Sup35MC fusion protein. Similar to the yeast containing wild-type Sup35, the yeast strain lacking the endogenous SUP35 gene, but harboring LEF-10-Sup35MC exhibited both [PSI + ]-like and [psi − ]-like phenotypes (Fig. 2a), which were hereafter referred to as [LEF + ] and [lef − ]. The replacement of Sup35 with the LEF-10-Sup35MC fusion proteins in [lef − ] cells gave rise to colonies with a red, Adephenotype indicating the LEF-10-Sup35MC fusion protein could efficiently terminate translation at the premature UGA stop codon in the ade1-14 allele in this strain. In contrast, but similar to [PSI + ] strains, [LEF + ] cells expressing the chimeric LEF-10-Sup35MC protein showed significantly reduced termination efficiency resulting in the read-through of ade1-14 premature stop codon and the white Ade + phenotype as seen in [PSI + ] cells. This behavior is consistent with the functional sequestration of LEF-10-Sup35MC into an aggregated state in [LEF + ] cells leading to reduced translation termination activity.
We further observed that [LEF + ] and [lef − ] strains could maintain the respective prion or non-prion phenotypes, i.e., suppression or non-suppression of the ade1-14 allele, over many cell generations. The white Ade + phenotype was stable during the propagation of [LEF + ] colonies, while [lef − ] strain spontaneously gave rise to [LEF + ] colonies at a very low frequency (Fig. 2a). As the SUP35 native promoter was used for the expression of the LEF-10-Sup35MC fusion protein, this observation suggested that the prion conformation of LEF-10 was self-perpetuating and the low expression level was sufficient for the maintenance of its prion state.
Most of prion proteins have the ability to form SDS-resistant polymers. The SDS resistance of protein complexes on semidenaturing detergent agarose gels (SDD-AGE) can distinguish highly ordered amyloid fibrils from disordered superstructures 30 . To assess whether LEF-10 possesses this typical SDS-resistant characteristic, we examined the LEF-10-Sup35MC fusion protein in [LEF + ] and [lef − ] strains using SDD-AGE. Consistent with the in vivo phenotypes, the LEF-10-Sup35MC protein formed highmolecular-weight, SDS-resistant aggregates in [LEF + ] yeast cells, but could only be detected as monomers in [lef − ] yeast cells (Fig. 2b). To exclude the possibility that different protein levels contributed to the formation of aggregation bands, the expression levels of chimeric LEF-10-Sup35MC proteins were tested in parallel by Western blot using SDS-PAGE (Fig. 2b, the bottom two panels). Detection of similar expression levels of these chimeric proteins suggested that the SDS-resistant characteristic harboring by LEF-10 was its innate biochemical property.
As a molecular chaperone, Hsp104 is indispensable for the heritability of yeast amyloid prions 31 . To test whether Hsp104 is essential for the maintenance of the [LEF + ] phenotype, the chaperone function of Hsp104 was inactivated using guanidine hydrochloride (GdnHCl), an inhibitor of the ATPase activity of Hsp104 32 . By growing [LEF + ] cells on medium containing 5 mM GdnHCl, [LEF + ] strains were restored to [lef − ] phenotype (Fig. 3a). The dependence of [LEF + ] on Hsp104 was further investigated by assessing the de novo inducibility of [LEF + ] in the absence of Hsp104. It was observed that deletion of the nonessential HSP104 gene eliminated the [LEF + ] phenotype and these cured [lef − ] strains only produced red Ade − colonies ( Fig. 3b) indicating that the LEF-10 aggregates required Hsp104 for its propagation and that the [LEF + ] phenotype in yeast was Hsp104-dependent. Such dependency is found for the majority of yeast prions 31,33 ruling out non-epigenetic effects and demonstrating the bona fide prion phenotype of [LEF + ] strains.
These results support our speculation that LEF-10 is a prionforming protein that not only can exist in at least one alternative self-perpetuating conformational state, but also was able to switch between a soluble form and a prion-like aggregated form in vivo. Identification of the cPrD in LEF-10. Almost all identified yeast prion proteins harbor glutamine/asparagine (Q/N)-rich domains, which are responsible for the prion characteristic of selfpropagating aggregation. LEF-10 is only 78 amino acids in length and does not contain any Q/N-rich regions. Using PLAAC, an algorithm used to predict PrDs in proteins based on their amino acid sequence 34 , we were unable to define a distinct candidate prion-forming domain (cPrD) in LEF-10 (Supplementary Fig. 2a). However, Tango, an algorithm used to predict amyloidogenic regions 35 , predicted four adjacent regions in the N-terminal part of LEF-10 as having potential of amyloid formation ( Supplementary Fig. 2b). To further define the potential cPrD, we generated three LEF-10 truncations based on the sequence alignment of LEF-10 in 28 baculoviruses 25 , with each truncation containing one of the three conserved regions (C1, C2, and C3) of LEF-10 and its flanking sequences (Fig. 4a).
In the Sup35MC-based in vivo assay, LEF-10 1-41 containing the C1 conserved region maintained [LEF + ] phenotype when fused to Sup35MC in yeast cells (Fig. 4b, c), consistent with the prediction by the Tango algorithm that a cPrD is located at the N terminus of LEF-10. We next shortened LEF-10 1-41 from each end and found that a 23 amino acid stretch (LEF-10 12-34 ) was able to drive the [LEF + ] phenotypes in yeast. Hence, we defined residues 12-34 as the minimal cPrD of LEF-10 ( Fig. 4a-c), although this sequence shows no sequence homology with any sequence known to function as a PrD in yeast.
Effects of highly conserved amino acids in the LEF-10 cPrD. Identification of the LEF-10 cPrD in the most conserved region of C1 prompted us to explore whether the conserved residues in LEF-10 cPrD were associated with the prion-forming property of this region. We therefore systematically substituted each of the ten highly conserved amino acid residues (Fig. 4a) in the identified cPrD with alanine, and tested whether these mutations could eliminate the [LEF + ] phenotype in yeast cells. None of these mutations completely eliminated the [LEF + ] phenotype or inhibited [LEF + ] propagation ( Supplementary Fig. 3), suggesting that the primary sequence of the LEF-10 cPrD is not a strict requirement for this property. We also investigated if the same Ala substitution mutants could replace the wild-type LEF-10 function when placed under the control of a strong promoter and expressed as fusion proteins with a C-terminal EGFP ( Supplementary Fig. 1b). Among the mutants, several mutants (LEF-10 I17A , LEF-10 N20A , and LEF-10 I29A ) lost the ability to rescue baculovirus using a lef-10 null bacmid and no late gene expression was detected, whereas the late gene expression level regulated by the mutant LEF-10 L21A was significantly higher than that of wild-type LEF-10 (Supplementary Fig. 4a and Fig. 5a, b). In order to determine the effect of L21A substitution on the function of LEF-10 under physiological conditions, baculoviruses were further rescued using wild-type LEF-10 or LEF-10 L21A expressed under the control of lef-10 native promoter (Supplementary Fig. 1d) and their growth curves were determined. Compared to the virus expressing wild-type LEF-10, the replication speed of the virus rescued by LEF-10 L21A decreased and its virus titer plateau was approximately 10-fold lower ( Supplementary Fig. 5), suggesting that LEF-10 L21A was a functionally down-regulated mutant.
As we had noticed that protein overexpression can induce the aggregation of LEF-10, we postulate that the aggregation tendency of LEF-10 L21A may be lower than the wild-type protein, and this property may compensate the down-regulated function of the mutant when overexpressed, and then upregulate its downstream gene expression. Using fluorescence microscopy, we observed that overexpressed LEF-10 L21A -EGFP formed many fewer aggregates than overexpressed LEF-10-EGFP in baculovirus-infected insect cells (Fig. 5c, d). In transformed yeast cells, LEF-10 L21A -GFP also produced fewer and smaller aggregates than LEF-10-GFP ( Supplementary  Fig. 6). These results are consistent with the L21A mutation reducing the aggregation tendency of LEF-10.
Prion behavior of LEF-10 in its native host. To investigate the prion property of LEF-10 in infected cells and the effects of protein aggregation on its function, we defined the two LEF-10 phenotypes in infected cells: LEF-10 existed as small punctate aggregates (LEF Ag ) in the cytoplasm; LEF-10 was diffused (LEF d ) and evenly distributed in the cell and/or localized in the nuclei (Fig. 6b). We counted the cells with the two phenotypes in Sf9 cells expressing LEF-10-EGFP driven by the lef-10 promoter ( Supplementary Fig. 1e), and found that the ratio of cells with LEF Ag continuously decreased from 24 to 48 hpi, whereas the ratio of cells with LEF Ag dropped more rapidly in cells expressing LEF-10 L21A than those expressing wild-type LEF-10 ( Fig. 5e, f). This observation is also consistent with our observation that the L21A mutation reduced the aggregation tendency of LEF-10 even at its normal expression level. SDD-AGE and Western blot analysis showed that the aggregation degree of the L21A mutant is lower than that of the wild type protein (Fig. 5g, h).
Using the p10 promoter driven EGFP as a reporter for baculovirus late gene expression ( Supplementary Fig. 1d), we infected insect cells at a multiplicity of infection (MOI) from 0.125 to 64 with the aim of controlling the expression and aggregation of LEF-10. At 36 hpi, we counted the cell numbers expressing EGFP. The results showed that the EGFP positive cells were comparable in LEF-10 L21A and wild-type LEF-10 groups at MOI lower than 8, however, at higher MOI, there were obviously more EGFP positive cells in LEF-10 L21A group than in the LEF-10 group ( Fig. 6a and Supplementary Fig. 7). We also measured the mean fluorescence intensity (MFI) of the EGFP positive cells at 60 hpi, and the data showed that the late gene expression levels were similar in the two groups at an MOI lower than 0.5, but MFIs for the LEF-10 L21A expressing cells were much higher than the cells expressing wild-type LEF-10 at an MOI higher than 1  Fig. 8). These results suggest that LEF-10 may be predisposed to switching to an aggregate state when an insect cell is infected by multiple virions, and the aggregation of LEF-10 will lead to the inactivation of the protein resulting in the reduction of the downstream gene expression that is dependent on it. Using the p10 promoter-driven mCherry as a reporter for baculovirus late gene expression ( Fig. 1a and Supplementary Fig. 1e), we found that for both the wild-type LEF-10 and the LEF-10 L21A mutant, the expression levels of mCherry in cells with LEF Ag were very low, but in contrast, all the cells expressing mCherry at high levels displayed cells with LEF d phenotype (Fig. 6b). These results demonstrated that LEF-10 can exist in two different conformations in virus-infected insect cells, and high-level expression of the protein induced the switch to the aggregate state which resulted in the inhibition of late gene expression.

Discussion
The data we present in this study shows that (a) LEF-10 can functionally replace the PrD of the yeast Sup35 protein allowing it to behave as a prion in yeast, and (b) that the LEF-10 protein can exist in two different conformations in insect cells. Such in vivo behavior is entirely consistent with LEF-10 being a baculovirusencoded prion-forming protein.
To date, all previously identified prions have been reported in cellular life forms. Whether viruses, the most abundant and diverse of life forms, might use prions was unknown and therefore our discovery extends prions to the viral world. Studies in yeast have revealed that many prion proteins contain high aggregation tendency domains that are commonly enriched for glutamine and asparagine residues. Based on these findings, several PrD prediction algorithms have been established 34,36-39 , which in turn have led to the discovery of more prion candidates with Q/N-rich domains 7,11,12 . Contrary to this common characteristic, functional PrDs have also been described in non-Q/Nrich prion proteins e.g., HET-s 9 , Mod5 8 and as shown in this report, LEF-10. These atypical PrDs do not share consensus primary sequences nor bias towards specific amino acid residues. As experimental identification of a prion protein is time consuming, especially on a genome-wide scale, development of a universal computational algorithm is a necessary task. This work is a challenge and partially depends on the discovery of more non-Q/N-rich prion proteins. Here, we provide another such instance that will inform future developments in these algorithms.
The majority of yeast prions are cytoplasmically-located and show non-Mendelian inheritance, however, virus-encoded prions are more likely to be associated with the regulation of virus replication due to the non-cellular nature of viruses. When baculoviruses infect insect cells at a high multiplicity and the expression level of LEF-10 reaches a threshold, the protein tends to form aggregates leading to a shutdown of the downstream gene expression. Such a regulatory mechanism may benefit the population transmission of baculovirus (Fig.6c). Surprisingly, being a sensor of this regulatory system, LEF-10 is more sensitive to form prion state than one might have expected. The blocking effect by LEF-10 only emerged at MOI of lower than 1 (Supplementary Fig. 8) and it is well established that a high MOI will reduce the production of recombinant protein using baculovirus as the expression platform 40 (although there is a contrary report for Orgyia pseudotsugata multicapsid nuclear polyhedrosis virus 41 ). Based on our findings, we postulate that LEF-10 is a selflimiting factor that blocks virus late gene expression at high MOI, and release of LEF-10-derived restriction might be an efficient way to improve the productivity of baculovirus expression vector system. The regulation of gene expression can occur at various levels. Altering protein activity through allosteric changes in a protein is no doubt a rapid way to regulate downstream gene expression, especially for viruses which have very short generation time. As parasitic organisms, the propagation of viruses depends on the environment provided by their hosts. The switchable conformation of a prion-forming protein could allow viruses to quickly respond to the stresses coming from their hosts. This advantage may encourage screening and retention of prion-like proteins during the rapid evolution of the virus. We further speculate that proteins with prion properties are unlikely to be rare in the virus world and more virus-encoded prion-forming protein would be discovered in the future.
Cross-seeding is a phenomenon whereby a heterologous aggregated protein can template the de novo conversion of another protein to switch to an aggregated form 29,42,43 . Several examples of human disease-related cross-seeding have been reported. For example, α-synuclein aggregates can promote the aggregation of Tau protein in neurons 42 . Amyloid beta (Aβ) can accelerate the aggregation of Tau in vivo 44 and it can also act as a seed to promote the aggregation of α-synuclein in vitro 45 . Furthermore, amyloid protein made by microbiota has been reported to enhance α-synuclein aggregation in aged rats and nematodes 46 . This accumulating evidence of widespread heterologous crossseeding in amyloid formation raises the possibility that pathogenderived aggregation-prone proteins may also trigger amyloid formation and hence disease in humans. Researchers have long noticed a correlation between viral infection and Alzheimer's disease 47 , and recent reports further confirmed this connection 48,49 . Although it is too early to generalize, with our discovery of the first virus-derived prion, more attention should now be focused on whether virus-encoded prions can trigger the biogenesis of cellular prion-related diseases following virus infection in animals, especially in humans.  Fig. 1b) and they could rescue the BacmidΔlef-10 (more mutants in Supplementary Fig. 4a). b Flow cytometry analysis revealed that the expression level of LEF-10 L21A -EGFP was higher than that of LEF-10-EGFP. c Laser confocal microscopy images of over-expressed LEF-10-EGFP and LEF-10 L21A -EGFP in infected Sf9 cells at 48 hpi (Related to Supplementary  Fig. 4c)  The lef-10 promoter and its coding sequence were cloned into this plasmid as a SacII-XbaI fragment and the construct was named as pTriEx-LEF-10/EGFP. Sitedirected mutagenesis of plasmids pTriEx-LEF-10-EGFP 25 , pTriEx-P lef-10 -LEF-10-EGFP 25 , pTriEx-LEF-10/EGFP, and pTriEx-P lef-10 -LEF-10-EGFP-P p10 -mCherry 25 were created with the Q5 ® Site-Directed Mutagenesis Kit (NEB).
To construct pUKC1620-derived plasmids (listed in Supplementary Table 1), varietal LEF-10 coding sequences were cloned into pUKC1620 50 Fig. 1e). Hoechst stains the nuclei. Scale bar, 50 μm. c Model of how the LEF-10 endows prion-mediated regulation on late genes, resulting in the limitation of viral replication fragment were cloned into plasmid p6431 as a SacII-NheI fragment in-frame with the GFP under the control of the yeast CUP1 promoter. All successfully cloned plasmids (Supplementary Table 1) were confirmed by sequencing. Primers sequences are listed in Supplementary Table 4.
Computational predication of protein aggregation propensity. The aggregation propensity predication for late expression factor LEF-10 (NP_054083.1) of Autographa californica multiple nucleopolyhedrovirus (AcMNPV) was analyzed by two different programs: (1) PLAAC 34  Plasmid shuffling and phenotypic analysis. Plasmids containing LEF-10-Sup35MC variants were, respectively, transformed into strain LJ14 by the standard PEG/LiAc/ssDNA transformation method 53 . The resulting transformants were incubated on histidine-deficient and uracil-deficient synthetic medium for 3 days at 30°C and then plated on YPD medium supplemented with 1 mg/ml of 5fluoroorotic acid (CAS: 703-95-7, Sangon Biotech) to eliminate the Sup35 maintainer plasmid p[SUP35-URA3], generating strains expressing a LEF-10-Sup35MC fusion protein as the only source of functional Sup35. Phenotypes were assessed by growth of colonies overnight at 30°C in YPD medium and streaking or spotting of different dilutions of these cultures on rich 1/4 YPD agar and rich 1/4 YPD agar with 5 mM guanidine hydrochloride 54 .
Post-shuffled yeast strains were plated on SD-Ade medium to verify the readthrough status of the premature UGA stop codon in the ade1-14 allele. In [PSI + ] cells, the majority of Sup35 aggregated and was unavailable for translation termination, resulting in the read-through of ade1-14 premature stop codon. The synthesis of full length Ade1 protein (N-succinyl-5-aminoimidazole-4carboxamide ribotide synthetase) resulted in white colonies on rich medium and growth on SD-Ade medium. In contrast, [psi -] cells showed no read-through of the ade1-14 premature stop codon and failed to grow on SD-Ade medium.
pUKC1620-LEF-10, a plasmid encoding a LEF-10-Sup35MC chimera substituting for full-length Sup35, was transformed in parallel with pUKC1620 containing the full-length SUP35 gene, into strain LJ14. Phenotypes of the postshuffled colonies were assessed as described above. [PSI + ]-like (i.e., white) and [psi -]-like (i.e., red) colonies were, respectively, passaged on YPD plates and re-streaked on SD-Ade medium to test the stability and maintainability of these two phenotypes.
Generation of an HSP104 gene deletion. A recombination-based PCR-generated deletion strategy 55 was used to delete the endogenous HSP104 gene in yeast. A replacement cassette containing the endogenous URA3 gene and its promoter fragment was amplified from p6431 using primers (listed in Supplementary Table 4), so that the replacement cassette was attached to two flanking homologous arms of the HSP104 sequence. The linear replacement fragment was transformed into recipient strains expressing either Sup35, LEF-10-Sup35MC, LEF-10 1-41 -Sup35MC, or LEF-10 12-34 -Sup35MC (listed in Supplementary Table 2) by standard PEG/LiAc/ssDNA transformation 53 . The resulting transformants were grown on SD-Ura medium for 3 days at 30°C and then plated or streaked onto 1/4 YPD and/or SD-Ade media.
Generation of recombinant baculoviruses. The recombinant baculoviruses in this research were generated by homologous recombination between BacmidΔlef-10 25 and pTriEx1.1-derived plasmids ( Supplementary Fig. 1) in co-transfected Sf9 cells using FuGENE ® HD Transfection Reagent (Promega). The baculoviruses supernatants were collected at 4-5 days post co-transfection and the supernatants were centrifuged at 500 × g for 5 min to remove cell debris. Then the baculoviruses titers were measured and infected Sf9 cells at specific MOI.
Flow cytometry. Sf9 cells were harvested at 36 hpi ( Fig. 6a and Supplementary  Fig. 7b), 48 hpi ( Fig. 5b and Supplementary Fig. 4a) or 60 hpi ( Supplementary  Fig. 8b, c) by centrifugation at 300 × g for 2 min, and the pellet was resuspended in PBS. Data detection and collection were performed on a CyFlow Cube 6 flow cytometer (PARTEC) at an excitation wavelength of 488 nm. Cells expressing EGFP fusion proteins were detected in the FL1 channel at an emission wavelength of 530 nm. Data were collected from at least 10,000 cells for each sample and cell populations were analyzed offline using FlowJo ® software. The infected Sf9 cells expressing EGFP (EGFP positive cells) were gated (Supplementary Fig. 7b) and EGFP positive cells were counted (Fig. 6a) at 36 hpi. Infected Sf9 cells were gated ( Supplementary Fig. 8b) and the mean fluorescence intensity (MFI) was calculated ( Supplementary Fig. 8c) at 60 hpi. Second Order Polynomial regression was performed for regression analysis using GraphPad Prism version 7.00 for Windows, GraphPad Software, La Jolla California USA, www.graphpad.com.
Cell imaging. The plasmid p6431-LEF-10 and p6431-LEF-10 L21A (Supplementary  Table 1) were individually transformed into the yeast strain LJ14-SUP1 (Supplementary Table 2) using the standard PEG/LiAc/ssDNA transformation method 53 . When yeast cells cultured in SD-Ura medium reach logarithmic growth phase, CuSO 4 was added to a final concentration of 25 μM to induce the expression of GFP fusion proteins. Yeast cells were imaged at 4 h post-induction with a Revolution XD spinning disk confocal system equipped with a CSU-W1 spinning-disk head (Yokogawa) and an iXon Ultra 888 EMCCD (Andor) on a DMi8 microscope body (Leica). Specifically, yeast cells were imaged with a HCX PL Apo 100 × /1.44 oil immersion objective (Leica). GFP was excited at 488 nm and a TR-F525/50 bandpass emission filter (Semrock Brightline) were used for capturing GFP signals. Images were captured with Andor iQ3 software (Andor). Confocal images were cropped and processed with the ImageJ Software.
Immunoblotting. To investigate the amyloid propensities of LEF-10, LJ14-derived yeast strains harboring various LEF-10-Sup35MC chimeras under the control of the endogenous SUP35 promoter were cultured to mid-exponential phase and then processed for semi-denaturing detergent-agarose gel electrophoresis (SDD-AGE) analysis 30 .
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All relevant data are available from authors upon request. The source data underlying Figs. 5d, 5f, 6a and Supplementary Figs. 5 and 8c are provided as a Source Data file.