Heterologous Aggregates Promote Prion Appearance via More than One Mechanism

Summary

Certain proteins can misfold into β-sheet-rich, self-seeding aggregates. Such proteins appear to be associated with neurodegenerative diseases such as prion, Alzheimer's and Parkinson's. Yeast prions also misfold into self-seeding aggregates and provide a good model to study how these rogue polymers first appear. De novo prion appearance can be made very frequent in yeast by transient overexpression of the prion protein in the presence of heterologous prions or prion-like aggregates. Here, we show that the aggregates of one such newly induced prion are initially formed in a dot-like structure near the vacuole. These dots then grow into rings at the periphery of the cell prior to becoming smaller rings surrounding the vacuole and maturing into the characteristic heritable prion tiny dots found throughout the cytoplasm. We found considerable colocalization of two heterologous prion/prion-like aggregates with the newly appearing prion protein aggregates, which is consistent with the prevalent model that existing prion aggregates can cross-seed the de novo aggregation of a heterologous prion protein. However, we failed to find any physical interaction between another heterologous aggregating protein and the newly appearing prion aggregates it stimulated to appear, which is inconsistent with cross-seeding.

Introduction

Prions were first described as self-perpetuating infectious agents devoid of nucleic acids that cause several fatal neurodegenerative diseases. Prion diseases, also known as transmissible spongiform encephalopathies (TSEs), were shown to infect a variety of mammals [1]. All known mammalian prion diseases are caused by conversion of largely α-helical cellular prion protein PrP^C into fibrous β-sheet-rich ordered aggregates (amyloids) called PrP^Sc (associated with scrapie) [2]. Curiously, PrP^Sc can exist in different heritable forms, called strains, which cause neurodegenerative diseases with different characteristics and pathologies [3]–[5].

A number of other neurodegenerative diseases are also associated with conversion of a soluble protein to amyloid. For example, amyloid-like forms of Aβ and Tau, α-synuclein, huntingtin, FUS/TLS, TDP-43 or SOD1 are linked respectively to Alzheimer's (AD) [6], Parkinson's (PD) [7], [8], Huntington's (Htt) [9] and Amyotrophic Lateral Sclerosis (ALS) diseases [10]–[15]. Factors that influence the spontaneous conversion to amyloid are of considerable interest as possible disease risk factors. One important finding is that heterologous amyloid can promote the de novo conversion of a protein to amyloid. For example, Aβ accelerated the in vivo aggregation of tau [16], and Aβ and α-synuclein seeded each other's aggregation in vitro [17]. Indeed, recently distinct conformational variants of α-synuclein aggregates were shown to differentially promote the aggregation of tau in neurons [18].

Several proteins in the simple eukaryote yeast have been shown to convert from soluble to amyloid. The amyloid forms of these proteins are self-propagating prions associated with transmissible phenotypes [19]–[28]. These proteins provide good models for the amyloid conversion of human disease proteins.

For both human and yeast proteins, only a portion of the protein, called the prion domain (PrD), converts to amyloid. This portion of the protein is required for prion induction and propagation [29], [30]. The PrD of most known yeast prions is glutamine (Q) and asparagine (N) rich. Likewise, several human aggregating disease proteins e.g. huntingtin, TDP-43 and FUS contain Q/N-rich regions [31], [32]. In contrast, the recently discovered yeast [MOD⁺] prion, composed of Mod5, as well as PrP, Aβ and α-synuclein do not contain Q/N rich domains [27]. Similar to the mammalian PrP strains, yeast prions can also fold into numerous heritable conformations, called variants, leading to different degrees of altered phenotypes [33]–[36].

The most well-studied yeast prion is [PSI⁺], the prion form of Sup35. In its native form, Sup35 (release factor 3) works with Sup45 (release factor 1) to promote translational termination at stop codons [37]. The Sup35 protein consists of three major domains: N-proximal (N domain) required for prion induction and propagation; a highly charged middle (M) domain conferring solubility to the molecule and containing Hsp104 binding sites [38] and a C-terminal (C) domain essential for translational termination and viability [33], [39]–[41]. [PSI⁺] forms when Sup35 molecules assemble into amyloid-like aggregates, causing loss-of-function in translation termination, which leads to read-through of stop codons [20], [42], [43].

The spontaneous appearance of prions in yeast is extremely rare. Indeed, the conversion of prion-free cells, [psi^-] to [PSI⁺] was determined to be ∼5.8×10⁻⁷ [44]–[47]. On the other hand, overproduction of full-length Sup35 or its prion containing domain (Sup35NM) can increase the de novo appearance of [PSI⁺] dramatically, presumably by increasing the chance of Sup35 prion domains to misfold and interact [48]. This enhanced formation of [PSI⁺] requires either the presence of another prion [22], [49], [50] or the simultaneous overexpression of heterologous Q/N-rich domains [22], [51]. The best studied example of this stimulation by a prion involves [PIN⁺], the prion form of the Rnq1 protein. Although [PIN⁺] dramatically enhances the appearance of [PSI⁺], it is not required for [PSI⁺] propagation [49].

Understanding how [PIN⁺] enhances de novo induction of [PSI⁺] will help us understand analogous interactions between heterologous human disease proteins. Several models have been proposed (reviewed in [52]). The cross-seeding model suggests that [PIN⁺] initially acts as a seed for the conversion of the Sup35 prion domain into the [PSI⁺] conformation. Once [PSI⁺] is established, it is proposed to create its own seeds independent of [PIN⁺], allowing it to continue to propagate efficiently [53]. Several in vitro studies provide evidence in favor of induction of [PSI⁺] via cross-seeding [52], [54], and for the enhanced rate of polymerization of other proteins in the presence of heterologous aggregates [55]–[61]. Most notably, mCherry:FUS fibers were extended in length heterotypically when exposed to monomeric GFP:hnRNPA1 [55]. Definitive evidence for cross-seeding in vivo is much more difficult to obtain. Still, a fusion of the prion domain of Sup35 (NM) and Rnq1 lead to the efficient induction of [PSI⁺] in the presence of [PIN⁺], even without Sup35 overexpression [62], presumably because the fusion efficiently brings Sup35NM to the Rnq1 aggregates, thereby increasing the chance of physical association and resulting cross-seeding. Also, different [PIN⁺] variants preferentially cause the genesis of different variants of [PSI⁺] [63], [64], which can be easily explained by cross-seeding but not by chaperone titration.

The titration model postulates that cellular factors responsible for the disassembly of aggregates and the refolding of misfolded proteins are so busy working on the existing [PIN⁺] prion that they are not available to prevent the appearance of the new prion, [PSI⁺] [22], [51], [65]. In support of this model, prion-like aggregates have been shown to colocalize with chaperones, reducing the cytosolic level of chaperones and thereby affecting the stability of heterologous prion aggregates in the cell [66]–[70].

Molecular chaperones, which are normally involved in proper protein folding play a critical role in the maintenance of yeast prions (reviewed in [71]). Particularly, the Hsp104 chaperone in conjunction with chaperones of the Hsp70 (Ssa/Ssb) and Hsp40 (Sis1) families was shown to shear prion aggregates into smaller fragments that promote fiber growth and transmission to daughter cells [72]–[77]. The shearing activity of Hsp104 is antagonized by millimolar concentrations of guanidine hydrochloride (GuHCl), leading to the loss of [PSI⁺] [78], [79], and other yeast prions [19], [50]. Hsp104 is required for the propagation of almost all known yeast prions [72], [80]–[82].

Stimulation of de novo generation of prions in yeast is achieved by inducing overexpression of the corresponding prion protein. The resulting aggregates have been monitored with fluorescent derivatives. The de novo induction of [PSI⁺] promoted by [PIN⁺] was shown to display various Sup35 aggregates and go through several stages. Overexpression of Sup35NM-GFP, gave rise to fluorescent dot, line and ring-like assemblies [83]–[85]. The fluorescent rings induced by Sup35NM-GFP overexpression is a hallmark of [PSI⁺] induction. Indeed, most viable ring/dot-bearing cells gave rise to [PSI⁺] progeny [83]–[86]. Sup35 dots appeared earlier than rings and lines [84]. Ring-like aggregates were shown to be first peripheral along the cell membrane, and later internal surrounding the vacuole [83], [85]. When cells with such rings were followed in media that turned down Sup35 overexpression, Sup35NM-GFP appeared as dots in daughter cells, a typical feature of [PSI⁺] [83]–[85]. Once [PSI⁺] is established, Sup35NM-GFP overexpression results in one or a few large mature dots, or clumps but rings do not appear at all [29], [43]. These large dots replace the numerous small Sup35-GFP aggregates seen in [PSI⁺] cells with endogenous Sup35 tagged with GFP prior to overexpression [75]. When Sup35NM-GFP was constitutively overproduced in [PIN⁺] cells with a deletion of the Sup35 prion domain, only internal rings were observed prior to the transition to mature dots [87].

In this study, we report that the de novo appearance of [PSI⁺] aggregates begins with dots that co-localize with the main Rnq1 aggregate near the vacuole, that grow into peripheral rings and lines prior to the appearance of internal rings. Our studies also reveal preferential colocalization of Rnq1 and Pin4C aggregates with newly appearing vs. established [PSI⁺] aggregates, which is consistent with the cross-seeding model for [PSI⁺] induction. However, the failure of Mod5 to physically interact with Sup35 during Mod5-promoted [PSI⁺] induction suggests that cross-seeding is not involved. Finally, we provide evidence for the [PIN⁺]-independent requirement of Hsp104 during [PSI⁺] induction in vivo.

Results

Newly induced Sup35 aggregates arise near the vacuole and then grow into peripheral rings

We used GFP-tagged SUP35 constructs to visualize the initial conversion of Sup35 from soluble to aggregated protein when [psi^-] cells were induced to become [PSI⁺] by overexpressing Sup35(NM). Sup35NM-GFP overexpressed in [PIN⁺][psi^-] cells (Fig. 1A) progressed over time from diffuse cytoplasmic fluorescence in all cells to some cells with one to three fluorescent foci one of which was always near the vacuole, to more cells with dots. Later, peripheral rings and lines started to replace dots in some cells (see Table 1 for details). In contrast, [pin^-] cells showed no aggregates of Sup35NM-GFP at any time point. Even when we dramatically reduced the level of Sup35NM-GFP overexpression by growing cells in 0.2 rather than 2% Gal, the type and order of appearance of these aggregates did not change (S1 Table).

In another approach, we examined endogenous Sup35 tagged with GFP when untagged Sup35NM was overexpressed in [PIN⁺][psi^-] cells (Fig. 1B, Table 1). Upon induction of Sup35NM, all cells initially showed diffuse Sup35-GFP fluorescence, which was later seen as cytoplasmic dots one of which was near the vacuole and then, lines and rings in some cells. Similar results were observed when untagged full-length Sup35 was overexpressed (S1A Fig.). As expected, [pin^-] cells always displayed diffuse Sup35-GFP molecules in the presence of Sup35NM overexpression (S1B Fig.).

Because the diffuse fluorescence of Sup35NM-GFP observed in the experiments above might have masked the visualization of initial Sup35 aggregates during [PSI⁺] induction, we overexpressed Sup35NM from Bimolecular Fluorescence Complementation [88], [89] (BiFC) constructs, Sup35NM-VN and Sup35NM-VC, simultaneously in [PIN⁺][psi^-] cells (Fig. 1C, Table 1). Prior to 8 h of induction, no fluorescence was detected. The lack of diffuse fluorescence suggests that Sup35 aggregation does not begin all over the cell. At 8 h, a few cells showed fluorescent dots near the vacuole revealed by FM4-64 staining [90], but no lines/rings were visible. By 24-55 h, peripheral rings and lines appeared and more cells displayed dots. In control [pin^-] cells, no fluorescence was detected.

To determine if the dots and rings/lines that appeared within 24 h of induction of Sup35NM-YFP overexpression show amyloid-like properties, we stained the newly appearing Sup35NM-YFP aggregates in [PIN⁺][psi^-] cells with Thioflavin T (ThT) (Fig. 2): 30% of the dots and 60% of the rings were ThT-positive. As expected, Sup35NM-YFP mature dots in [pin^-][PSI⁺] were all stained with ThT. In control [pin^-][psi^-] cells however, diffuse Sup35NM-YFP fluorescence did not show any ThT fluorescence. These data suggested that Sup35NM-YFP does not always form amyloidogenic aggregates during [PSI⁺] induction, but eventually becomes amyloid in mature [PSI⁺].

To further investigate the subcellular localization of the initial Sup35 dots, we used BY4741 cells with genomic HSP42 tagged with GFP [91] (see Table 2 for details). Hsp42 is a small heat shock protein that appears as one big dot near the vacuole, sometimes referred to as the IPOD for the site(s) of deposit of insoluble protein aggregates [92]–[94]. Overexpression of Sup35NM-RFP in [PIN⁺][psi^-] HSP42-GFP cells first caused the occasional appearance of cells with 1-6 dots, one of which always colocalized with the Hsp42-GFP dot (Fig. 3). Later, in some cells, Sup35NM-RFP fluorescence extended from a bright dot that colocalized with the Hsp42-GFP dot as short lines tangent to the vacuole or as lines extending to the cell periphery. Interestingly, the multiple Sup35NM-RFP dots observed initially were never seen later once lines appeared, suggesting that Sup35NM-RFP aggregates that did not colocalize with Hsp42-GFP were solubilized, or may have joined the lines. Eventually, in some cells, Sup35 formed internal rings surrounding the vacuole as seen previously [83], [84], intersecting the Hsp42-GFP dot, and in a very few cells, lines were seen to extend from the Hsp42-GFP dot peripherally and around the vacuole simultaneously.

To determine the localization of Sup35 newly induced aggregates with respect to the vacuole, we overexpressed Sup35NM-RFP in [PIN⁺] cells with genomic VPH1 tagged with GFP (S2 Fig.). Vph1 is a subunit of the vacuolar-ATPase protein and marks the vacuolar membrane [95]. We found that Sup35 early dots (after 24 h of Sup35NM-RFP overexpression) were localized near the vacuole, and later, short lines extended outward from the vacuole to the periphery of the cell. Then, as expected, Sup35 formed peripheral rings, and eventually perivacuolar rings.

In summary, the various experiments above showed that during the de novo aggregation of Sup35 induced by its overexpression, Sup35 initially formed dots, one of which perfectly colocalized with the Hsp42-GFP dot near the vacuole. Then, Sup35 lines extended from this dot to form peripheral and eventually perivacuolar rings, while the other initial Sup35 dots disappeared.

Colocalization of Rnq1 with Sup35 dots

To visualize the relationship of Sup35 and Rnq1 during the de novo induction of [PSI⁺], we expressed Rnq1-GFP under its own promoter, and overproduced Sup35-RFP in [PIN⁺][psi^-] cells (Fig. 4A, top). Sup35-RFP initially formed fluorescent dots but no lines or rings (Details in S2 Table). All the Rnq1-GFP dots perfectly colocalized with Sup35-RFP dots, but only 60% of the Sup35-RFP dots colocalized with Rnq1-GFP foci. Interestingly, the colocalized Rnq1-Sup35 dots were always associated with the vacuole revealed by FM4-64 staining (Fig. 4A, middle). Additional Sup35 dots that did not overlap Rnq1 were away from the vacuole (Fig. 4A, gray arrows).

These Sup35 dots have different characteristics from Sup35 dots seen in mature [PSI⁺] cells. For example, while the results above showed that newly appearing vacuole-associated Sup35 aggregates perfectly colocalize with Rnq1, mature Sup35 aggregates in established [PSI⁺] cells did not entirely overlap Rnq1 (seen as in two sets intersecting in a Venn diagram) (Fig. 4A, bottom). Also, additional Rnq1-CFP dots existed in [PIN⁺][PSI⁺] cells that did not show any colocalization with Sup35 (Fig. 4A bottom, arrows, enlarged box).

In summary, after 6 h of [PIN⁺]-promoted Sup35-RFP aggregation, all Rnq1 dots perfectly overlapped newly induced Sup35 dots around the vacuole, but additional Sup35 dots (away from the vacuole) without overlapping Rnq1 sometimes existed in those cells. In contrast, in established [PSI⁺] cells, all Sup35 dots partially overlapped Rnq1 dots, but additional Rnq1 dots without overlapping Sup35 also existed.

Colocalization of Rnq1 with Sup35 rings

To look for colocalization of Rnq1 with Sup35 rings, we induced overexpression of Sup35-RFP for 24 h in [PIN⁺] cells also expressing Rnq1-GFP from its own promoter. Sup35-RFP formed rings 70% of which colocalized with Rnq1-GFP (Fig. 4B). In the remaining 30% of cells with Sup35-RFP rings that did not colocalize, Rnq1-GFP fluorescence was instead diffuse or in the form of dots (S3A Fig., S3 Table). In contrast, essentially all Rnq1-GFP rings colocalized with Sup35-RFP rings (S4 Table). As expected, Rnq1-GFP and Sup35-RFP always remained diffuse in [pin^-] cells (S3B Fig.).

In another version of this experiment, Sup35NM-YFP was overexpressed in [PIN⁺][psi^-] cells expressing Rnq1-CFP under its own promoter. Sup35NM-YFP overexpressed for 24 h formed fluorescent dots or rings respectively, in 7 and 0.8% of the cells. In these cells, 90% of Sup35NM-YFP dots colocalized with Rnq1-CFP dots, and all Sup35NM-YFP rings colocalized with Rnq1-CFP rings (Fig. 4C top, S5 Table). Curiously, these Rnq1-CFP rings looked like beads on a string, rather than an uninterrupted full ring.

After 48 h of Sup35NM-YFP overexpression, Sup35NM-YFP formed rings in 11% of the cells, and 75% of these Sup35NM-YFP rings colocalized with Rnq1-CFP rings (Fig. 4C bottom). In the remaining 25% of cells with Sup35NM-YFP rings, Rnq1-CFP showed diffuse or dot fluorescence (S6 Table). On the other hand, essentially all Rnq1-CFP rings colocalized with Sup35NM-YFP rings (S7 Table). In control experiments, [pin^-] cells never formed any aggregates when Sup35NM-YFP was overexpressed (S4A Fig.). [PIN⁺] cells showed only Rnq1-CFP dots, but no lines/rings when Sup35NM-YFP expression remained repressed in 2% Glucose (S4B Fig.) and when cells with empty vector expressing YFP were grown in 2% Gal (S4C Fig.). These colocalization data were based on visually checking different planes of the cells by moving the focal plane up and down, and were also confirmed by collecting z-stacks from representative cells (S5 Fig.).

Next, we co-overexpressed Rnq1-YFP and Sup35NM-VN in [PIN⁺][psi^-] cells (Fig. 4D). In 24 h, 7% of the cells had one Rnq1-YFP dot with lines extending from it in all directions, referred to as mesh-like aggregates. In controls, when Rnq1-YFP was overexpressed in [PIN⁺] without overexpressing Sup35NM, 90% of the cells showed multiple fluorescent dots, 10% had diffuse fluorescence, and none had mesh-like aggregates. Also, overexpressing Rnq1-YFP and Sup35NM-VN simultaneously in [pin^-] control cells did not result in any aggregate formation. When cells from cultures with Rnq1-YFP mesh-like aggregates (7%) were scored for [PSI⁺], 6.5% of these cells formed pink or white colonies and were able to grow on media lacking adenine (SD-Ade), indicative of [PSI⁺] (See Methods).

To visualize if Sup35 and Rnq1 have a close physical interaction during [PSI⁺] induction, we co-overexpressed Sup35NM-VN and Rnq1-VC in [PIN⁺][psi^-] cells (S6 Fig., S8 Table). Initially (16 h post induction) 1.8% of the cells showed dots, but no lines/rings; but later (40 h post induction) peripheral rings, lines and mesh-like aggregates appeared and more cells displayed dots. We also co-overexpressed Sup35NM-VN and Rnq1-VC in established [PSI⁺] as well as in [pin^-][psi^-] cells and did not observe any fluorescence. These data indicate that Rnq1 and Sup35 interact in a close proximity during [PSI⁺] induction, but not in established [PSI⁺].

As expected, overexpression of untagged Sup35NM caused Rnq1-CFP expressed from its own promoter to align in ring/line-like aggregates in 10% of the [PIN⁺] cells (Fig. 4E). The Rnq1-CFP lines looked like beads on a string as seen previously in the presence of overexpressed YFP-tagged Sup35NM (Fig. 4C top). Such beads on a string never appeared in [PIN⁺] cells without overexpressed Sup35NM, where Rnq1-CFP fluorescent dots remained dispersed, or in [pin^-], where Rnq1-CFP remained diffuse (S7 Fig.). When cells from the culture with Rnq1-CFP beads on a string (10%) were scored for [PSI⁺], 11% of these cells formed pink or white colonies, indicative of [PSI⁺].

The above experiments show that during the de novo aggregation of Sup35 in [PIN⁺] cells, overexpression of Sup35NM induced Rnq1 to form mesh-like or line/ring-like aggregates. Essentially all Rnq1 line/rings perfectly overlapped Sup35 line/rings, while some Sup35 line/rings did not overlap Rnq1. BiFC between Rnq1 and Sup35 confirmed that they form a close physical interaction during the induction of [PSI⁺], and initially form dots, and then lines/meshes; while they do not form such an interaction in established [PSI⁺].

Localization of other proteins during [PSI⁺] induction

In order to determine what other proteins (particularly chaperones) colocalize with Sup35 during its aggregation in the presence of [PIN⁺], we overexpressed Sup35NM-RFP in [PIN⁺] cells with endogenously tagged GFP proteins [91]. As seen previously [96], we found that the molecular chaperones Hsp104, Sis1, and Ssa1 involved in [PSI⁺] propagation via their involvement in prion shearing [72], [97]–[104], colocalized with Sup35 rings and dots perfectly during de novo induction of [PSI⁺] (Fig. 5A, S8 Fig.). However, Ydj1, a Hsp40 co-chaperone shown to co-immunoprecipitate with [PSI⁺] aggregates as a minor component along with Hsp104, Ssb, Sis1, Sse1 [104], [105] did not colocalize with Sup35 rings (Fig. 5A).

We also asked if proteins other than chaperones, previously found to influence the maintenance or induction of [PSI⁺] [22], [106], [107], would colocalize with newly appearing Sup35 aggregates. We found that Sup45-GFP perfectly colocalized with newly appearing Sup35NM-RFP aggregates, as well as with established [PSI⁺] aggregates (Fig. 5B). Upon testing other candidate proteins (Cyc8, New1, Pin3, Pin4, Tup1, Mod5, Sgt2), we found that none of them displayed any colocalization with Sup35NM-RFP aggregates (S9 Fig.).

Colocalization of Q/N-rich Pin4C vs. non-Q/N rich Mod5 with newly appearing Sup35 aggregates during [PSI⁺] induction in the absence of Rnq1

Next, we investigated if high levels of Pin4C (120–668 a.a.), which were previously shown to substitute for [PIN⁺] in promoting de novo induction of [PSI⁺] [22], [66], would overlap Sup35 aggregates during [PSI⁺] induction in the absence of Rnq1. We simultaneously overexpressed Pin4C-RFP and Sup35NM-GFP in 74D-694 rnq1Δ [psi^-] cells (Fig. 6, Table 3). Both proteins initially remained diffuse, but by 8 h, tiny Pin4C-RFP fluorescent dots appeared in some of the cells, all of which still had diffuse fluorescence of Sup35NM-GFP. At 16 h, Sup35NM-GFP fluorescent dots appeared and essentially all of these colocalized with Pin4C-RFP near the vacuole. At 24 h, Sup35NM-GFP started to appear as fluorescent rings in addition to dots. In these cells, essentially all Sup35NM-GFP aggregates overlapped Pin4C-RFP. At 48 h, the number of cells with Sup35NM-GFP dots decreased, while cells with rings increased, and all Sup35 aggregates still overlapped Pin4C-RFP. At 72 h, the Sup35NM-GFP dots still colocalized with Pin4C-RFP. However, almost all of the cells that had Sup35NM-GFP rings failed to show Pin4C-RFP rings. Instead, they contained large fluorescence spots of Pin4C-RFP that did not colocalize with Sup35NM-GFP. In control rnq1Δ cells that separately overexpressed either Sup35NM-GFP or Pin4C-RFP for 72 h, respectively, Sup35NM-GFP always remained diffuse, while in 50% of the cells (n≈450) Pin4C-RFP formed large fluorescent dots (S10A Fig.). As expected, [PSI⁺] appeared de novo in rnq1Δ cultures with overexpressed Sup35NM-GFP and Pin4C-RFP (S9 Table).

To determine the location of Pin4C aggregates relative to the vacuole during Sup35 de novo aggregation, we overexpressed Pin4C-RFP in [pin^-] HSP42-GFP cells in the presence vs. absence of Sup35NM overexpression (S10B Fig.). In the presence of Sup35NM overexpression, Pin4C-RFP formed one to a few foci, one of which perfectly (in 6.5% of the cells; n∼800) or partially (in 93.5% of the cells) overlapped the Hsp42-GFP dot. In the absence of Sup35NM overexpression, Pin4C-RFP dots never perfectly overlapped the Hsp42-GFP dot; rather one of the Pin4C-RFP aggregates was juxtaposed to, or partially colocalized with the Hsp42-GFP dot. This suggests that Sup35NM overexpression promotes a more frequent closer association of the Pin4C-RFP aggregate and the vacuole-associated Hsp42-GFP protein deposit.

When we simultaneously overexpressed the non-Q/N rich prion protein Mod5 tagged with GFP and Sup35NM-RFP in 74D-694 rnq1Δ cells, Sup35NM-RFP initially formed dots, and then rings/lines, while Mod5-GFP always remained diffuse (Fig. 7A, Table 4). In control rnq1Δ cells that separately overexpressed either Sup35NM-RFP or Mod5-GFP, both proteins remained diffuse (S11A Fig.). Since we could not see fluorescent aggregates of Mod5-GFP, we turned to BiFC to look for an interaction between Mod5 and Sup35NM (Fig. 7B, S11B Fig.). Overexpression of Mod5-VN and Sup35NM-VC did not result in any fluorescence in [pin^-] cells, although [PSI⁺] was induced in this culture with a frequency of 0.9% (n∼1000) suggesting that cross-seeding may not be universal for [PSI⁺] induction. Curiously, in [PIN⁺] cells, overexpression of Mod5-VN and Sup35NM-VC resulted in first diffuse fluorescence and then the formation of a single (near the vacuole) to multiple dots over time in 15% of cells (Fig. 7B). Also, Mod5-VN and Mod5-VC overexpression in [pin^-] cells did not show any fluorescence, but in [PIN⁺] cells they showed diffuse fluorescence (S11C Fig.). Possibly, newly appearing Sup35NM aggregates in [PIN⁺] cells are attracted to the Mod5 aggregates seen as diffuse fluorescence.

These findings indicate that during the de novo aggregation of overexpressed Sup35 promoted by overexpression of Q/N-rich Pin4C in the absence of Rnq1 [22], [66], Sup35 aggregates initially colocalize with Pin4C aggregates (near the vacuole), but Pin4C falls off the Sup35 rings later. Furthermore, Pin4C-RFP perfectly overlapped Hsp42-GFP only in the presence of Sup35 overexpression. The data is consistent with the cross-seeding of Sup35 aggregation by Pin4C aggregates in the absence of Rnq1. However, although the overexpression of the non-Q/N rich protein Mod5 promotes [PSI⁺] induction [27], we could not visualize Mod5-Sup35 direct interaction in the absence of [PIN⁺] suggesting that Mod5-promoted de novo Sup35 aggregation occurs via a different mechanism.

The role of Hsp104 in de novo [PSI⁺] induction

The Hsp104 chaperone requirement for the maintenance of [PSI⁺] [72] and the colocalization of Hsp104 with Sup35 dots and rings during the induction of [PSI⁺] ([96], Fig. 5A) led us to ask if Hsp104 is also required for the de novo aggregation of Sup35 during [PSI⁺] induction. Since Hsp104 is required for the maintenance of [PIN⁺], and the requirement for [PIN⁺] in the de novo induction of [PSI⁺] can be overcome by overexpressing certain Sup35NM-containing fragments, e.g. with a short extension of hydrophobic residues [108], we overexpressed Sup35NM with a short extension of hydrophobic residues (magic tail), previously shown to induce [PSI⁺] even in [pin^-] cells [49], [108]. Overexpression of Sup35NM with this magic tail (Sup35NM-mt) in [pin^-] HSP104 cells, caused endogenous Sup35-GFP molecules to form dots and short lines in 6% of the cells (n≈1200) (Fig. 8A) and this increased to 15% (n≈1400) of the cells in the presence of [PIN⁺]. Unlike the Sup35 dots induced in [PIN⁺] cells by 24 h of overexpression of Sup35NM without magic tail, Sup35 dots induced by Sup35NM-mt in [pin^-] HSP104 or [PIN⁺] HSP104 cells did not always appear near the vacuole. In [pin^-] hsp104Δ cells however, no dots were seen and only 0.4% of the cells (n≈1700) formed Sup35-GFP lines. Furthermore, all these lines appeared to be at the cell membrane as opposed to those seen in the cytoplasm in HSP104 cells.

In an alternative approach to test the role of Hsp104 in [PSI⁺] induction, we used Pin4C overexpression to substitute for [PIN⁺]. We co-overexpressed Sup35NM-GFP and Pin4C-RFP in [pin^-] hsp104Δ cells (Fig. 8B), and Sup35NM-GFP remained diffuse, while in control [pin^-] HSP104 or [PIN⁺] HSP104 cells, as expected, Sup35NM-GFP formed dots and rings/lines respectively in 7% and 9% of the cells. Also, Pin4C-RFP aggregates were not affected by the presence vs. absence of Hsp104 in [pin^-] cells, but were larger in the presence of [PIN⁺] (S12 Fig.). Co-overexpression of Sup35NM-GFP and Pin4C-RFP in [pin^-] HSP104 cells in the presence of GuHCl, which inhibits Hsp104's ATPase activity [79] resulted in the formation of only Sup35NM-GFP dots in 2.8% of cells, but no rings/lines and failed to induce any [PSI⁺] cells (Fig. 8C). Taken together, the striking differences in the level and types of Sup35 aggregate formation in the presence vs. absence of Hsp104 and in the presence vs. absence of GuHCl suggest that Hsp104 is required for the formation of de novo Sup35 aggregates, and induction of [PSI⁺] de novo.

Discussion

Protein aggregates have been implicated in a wide variety of diseases including Amyotrophic Lateral Sclerosis, Alzheimer's, Parkinson's and prion disease [109], [110]. Interactions between proteins associated with protein misfolding diseases (PMD) are of great interest since molecular cross-talk between disease aggregates of one protein can accelerate the de novo appearance of heterologous disease protein aggregates [16], [60], [111]–[115]. Several reports implicate cross-seeding as a mechanism to explain this [17], [18]. Here, our data provide insight into the mechanism of prion induction, which is a model for such heterologous interactions in human diseases. Our data suggest that cross-seeding is not the only mechanism for this cross-talk phenomenon. While our findings that [PIN⁺] or Pin4C aggregates physically interact with Sup35 de novo aggregates are consistent with the cross-seeding model, the failure of Mod5 to interact with Sup35 suggests the existence of another mechanism, possibly via hindering the chaperone network.

We stimulated the de novo generation of a prion in yeast by overexpressing fluorescent derivatives of the prion protein and then monitored its aggregation status. Based on our findings (Fig. 1–3, 6, S1–2 Fig.), we propose a model to explain the pathway followed by Sup35 during its de novo conversion into mature [PSI⁺] (Fig. 9A). Bimolecular complementation [88], [89] using BiFC-tagged Sup35NM (Fig. 1C) showed that Sup35NM molecules do not associate throughout the cytoplasm, but first interact at specific sites in the cell. This means that the de novo aggregation of Sup35 first occurs at these sites rather than being brought to them as pre-existing aggregates. This may be true of other prion-like aggregating proteins as well.

How could prion-like proteins first interact within discrete inclusions? We speculate that upon initial expression, Sup35 is soluble, but upon overexpression, some molecules of the intrinsically unstructured Sup35NM misfold. We suggest that this misfolded protein could be captured by quality control compartments (QCCs) as inclusions, i.e. recently discovered Q-bodies [116], where misfolded proteins accumulate en route to degradation. The high local concentration of Sup35 at these sites increases the likelihood of prion induction. Furthermore, [PIN⁺] prion aggregates, which are required for [PSI⁺] induction [21], [22], [51], are also located at these sites, where they could facilitate nucleation for Sup35 [53], [87], [117], [118] to polymerize Sup35 into amyloid. In Pin4C-promoted [PSI⁺] induction, Pin4C appears to take over the role of [PIN⁺] aggregates. This suggests that the co-existence of misfolded protein with heterologous amyloid in inclusions accelerates de novo conversion of the misfolded protein into amyloid.

However, not all of these inclusions give rise to larger Sup35 fibrils. Rather, Sup35 aggregates away from the perivacuolar site disappear, and Sup35 lines and rings emanate only from the single inclusion in the cell near the vacuole (Fig. 1–3,6). This perivacuolar inclusion also differs from the other inclusions because it alone colocalizes with Hsp42. Furthermore, the finding of only one fluorescent dot in [PIN⁺] cells with overexpressed BiFC-tagged Sup35NM and Rnq1 (S6 Fig.) suggests that Sup35 and Rnq1 interact only at the perivacuolar site. It is unclear what happens to the other inclusions, they could be solubilized, degraded or join the remaining perivacuolar aggregate. This suggests that the fibrillar growth of de novo aggregates requires site-specific chaperones.

Curiously, heterologous aggregates are not only involved in the initial cross-seeding, but continue to be associated with some newly seeded heterologous fibrils. This is surprising since once nucleated by [PIN⁺]/Pin4C amyloid, Sup35 polymerization should continue without the need for Rnq1/Pin4C seed. Indeed, due to the higher efficiency of homotypic polymerization [55], [118], [PIN⁺]/Pin4C aggregates are not expected to incorporate into the growing Sup35 fibrils. Paradoxically, Rnq1 frequently (Pin4C always) are found to overlap newly appearing Sup35 rings (Fig. 4, 6). We propose that although the initial step in nucleation is heterotypical, [PIN⁺]/Pin4C aggregates might also grow in length homotypically in close proximity to the Sup35 aggregates via a lateral interaction, possibly with the help of chaperones that are associated with the aggregates. Indeed, our findings that all Rnq1/Pin4C rings colocalized with Sup35 rings (S4 Table); and that Rnq1/Pin4C formed mesh/ring-like aggregates only in the presence of Sup35 overexpression (Fig. 4D–E, S6 Fig.) support this hypothesis, which predicts that Sup35 rings template the continued growth of Rnq1/Pin4C into rings and not vice versa (see Fig. 9B).

Mature [PSI⁺] dots do not associate with [PIN⁺] in the same manner as newly induced [PSI⁺] aggregates do. Indeed, although Rnq1 was found to co-immunoprecipitate with newly appearing Sup35 aggregates [63], [119], it was not detected in purified [PSI⁺] aggregates [104]. Here, we show the formation of fluorescent dots and mesh-like aggregates by co-overexpressed Rnq1-VC and Sup35NM-VN during [PSI⁺] induction, and the inability of Rnq1-VC and Sup35NM-VN to cause fluorescence in established [PSI⁺], suggesting that Rnq1 and Sup35 are in close proximity during [PSI⁺] induction, but not so close in established [PSI⁺] (S6 Fig., S8 Table).

Mature and newly induced Sup35 aggregates also differ in their amyloid characteristics as our data showed that all of the mature [PSI⁺] aggregates were stained in situ with amyloid-binding dye, thioflavin T, while only 30% and 60% of respectively, newly appearing Sup35 dots and rings were stained with ThT (Fig. 2). This could be either because cross-talk between heterologous amyloid aggregates may not always convert the prion-like protein into amyloid vs. amorphous aggregates, or interference of other proteins attracted to aggregates in situ may cause false negative results. We favor the latter possibility as most viable Sup35 ring-bearing cells give rise to [PSI⁺] progeny [84], [85]. However, it is possible that dead cells have non-amyloid Sup35 rings. Alternatively, it is also possible that newly appearing Sup35 aggregates that are stained with Thioflavin T harbor Rnq1 amyloid, leading to the ThT staining, while those Sup35 aggregates that are not stained do not harbor Rnq1 amyloid, leading to the failure of ThT staining. This would imply that the Sup35 molecules in the dot and ring structures are not amyloid.

While colocalization of a protein with heterologous aggregates is consistent with cross-seeding, it is not proof of cross-seeding. Indeed, the strict and permanent colocalization of Sup45 with Sup35 newly appearing aggregates and with mature [PSI⁺] aggregates (Fig. 5B) suggests that Sup45-GFP simply decorates all Sup35 aggregates. However, the considerable and transient colocalization of respectively, Rnq1 and Pin4C with Sup35 newly appearing aggregates, but only partial and no colocalization of respectively, Rnq1 and Pin4C [66] with mature [PSI⁺] aggregates supports the idea that as opposed to Sup45, Rnq1 and Pin4C actually cross-seed Sup35 de novo aggregates. It is noteworthy that Sup45 is not required for [PSI⁺] induction [86], but [PIN⁺], or one of its substitutes, Pin4C is.

Also, the colocalization of the molecular chaperone Hsp104 with all Sup35 aggregates (Fig. 5A, [120]) suggests that Hsp104 decorates rather than cross-seeds Sup35. Such decoration could enable Hsp104 to perform its known shearing activity of [PSI⁺] aggregates that is required for [PSI⁺] propagation [72], [80]–[82]. We showed that Hsp104 itself is also required for de novo induction of [PSI⁺] by finding inhibition of the de novo aggregation of Sup35 in [pin^-] cells lacking Hsp104 (Fig. 8A,B) and inhibition of de novo induction of [PSI⁺] in [PIN⁺] cells in the presence of GuHCl (Fig. 8C), which inhibits Hsp104's activity [79]. This is consistent with a previous report that overexpression of Hsp104 enhances prion appearance [121].

Although cross-seeding is generally thought to explain the cross-talk that enables amyloid aggregates to promote conversion of heterologous prion-like protein to amyloid, our data suggest that another mechanism is also involved. Indeed, we showed that in the absence of [PIN⁺], overexpression of the non-Q/N rich prion protein Mod5 enhances [PSI⁺] formation without direct physical interaction with Sup35 (Fig. 7B, S11B Fig.). Possibly, instead of cross-seeding which requires a physical interaction, excessive amounts of misfolded Mod5 proteins in [pin^-] cells sequester chaperones away from newly forming Sup35 aggregates in the cell, and thus allow them to mature into a prion. Since Mod5-promoted [PSI⁺] induction is rare compared to [PIN⁺]-promoted induction, cross-seeding appears to be more efficient than the other mechanism in promoting de novo prion formation.

[PIN⁺] also appears to influence de novo aggregation of Mod5. Bimolecular complementation using BiFC-tagged Mod5 (S11C Fig.) showed that Mod5 prion protein interacts with itself, resulting in diffuse fluorescence only in [PIN⁺], but not in [pin^-] cells. Curiously, an interaction between Sup35NM-VC and Mod5-VN is also seen as diffuse fluorescence, and again only in [PIN⁺] cells (Fig. 7B). Possibly, the Mod5 diffuse aggregates present in [PIN⁺] cells attract and interact with Sup35NM aggregates.

The data presented here aid our understanding of how prion formation occurs in yeast, and provide clues to the molecular mechanisms underlying many human aggregating neurodegenerative diseases, particularly these arising more frequently in people with preexisting neurodegenerative disease.

Methods

Strains and plasmids

Yeast plasmids and strains used in this study are listed in Tables 5 and 6, respectively. All [PIN⁺] cells used in the study were high [PIN⁺] [34]. GF657, GF658 and GF844 are, respectively, [PSI⁺][PIN⁺] HSP104, [psi^-][pin^-] HSP104 and [psi^-][pin^-] hsp104Δ versions of 74-D694 with endogenous SUP35 replaced with SUP35-GFP (kindly supplied as SY80, SY84, and SY97 by T. R. Serio, U. Arizona) [75]. L3107 and L2903 were obtained independently by curing GF657 of [PSI⁺] by overexpressing Pin4C [66]. GF852 and GF855, respectively, are [psi^-][PIN⁺] and [psi^-][pin^-] versions of 74-D694 with RNQ1-CFP under its own promoter integrated into the genomic TRP1 (kindly supplied as 645 and 651 by L. Li, Northwestern U.). GF647 was constructed by replacing chromosomal RNQ1 with KanMX4 in 74D-694 [122]. All other yeast strains used in the studies of colocalization with Sup35 are from the BY4741 GFP library strain (Life Technologies, CA) harboring the gene of interest tagged endogenously with GFP [91]. To obtain [pin^-] versions of these cells, they were grown on YPD plates with 5 mM GuHCl for three passes [50].

All overexpression plasmids in this study were driven by the GAL1 promoter unless otherwise stated. p1951 (Sup35NM-GFP) was constructed by inserting the SUP35NM BamHI-NotI fragment, and the GFP NotI-SacI fragment in-frame into the pRS413GAL1 plasmid backbone. p1893 (Sup35NM-VN) was constructed by replacing the GFP NotI-SacI fragment in p1951 with the VN173 NotI-SacI fragment PCR-amplified from pFA6a-HIS3MX6-pGAL1-VN173 [88]. p1893-2 was constructed by moving the GAL1-SUP35NM-VN XhoI-SacI fragment in p1893 to the pRS415 backbone. p2170 was constructed by replacing the SUP35NM BamHI-NotI fragment in p1893 with the MOD5 BamHI-NotI fragment PCR-amplified from the genome. p2171 was constructed by replacing the SUP35NM BamHI-NotI fragment in p1892 with the MOD5 BamHI-NotI fragment PCR-amplified from the genome. p1892 (Sup35NM-VC) was constructed by (1) replacing the GFP NotI-SacI fragment in p1951 with the VC155 NotI-SacI PCR-amplified from pFA6a-HIS3MX6-pGAL1-VC155 [88], and (2) moving this GAL1-SUP35NM-VC155 XhoI-SacI fragment into the XhoI-SacI sites of pRS414. p1894 was constructed by replacing the SUP35NM BamHI-NotI fragment in p1892 with the RNQ1 BamHI-NotI fragment PCR-amplified from the genome. p1984 was constructed by cloning the XmaI-Sal1 fragment of p1156 into p742. p2036 was constructed by replacing the YFP SpeI-XhoI fragment in p1752 with the NM(TAA) SpeI-XhoI fragment PCR-amplified from the genome [122]–[124]. p1753 was described previously [87], [96], [125], [126]. p2017 and p2018 vectors (Sup35NM-RFP) with URA3 or LEU2 markers, respectively were constructed in a two-step cloning: (1) The GFP NotI-SacI fragment in p1951 was replaced with the RFP NotI-SacI fragment amplified from p1708. (2) Then, this GAL1-NM-RFP XhoI-SacI fragment was cloned into pRS416 (URA3) and pRS415 (LEU2) backbones, respectively. All other plasmids listed in Table 5 were described previously [49], [62], [66], [85], [122], [127].

Cultivation procedures

Yeast strains were cultivated using standard media and growth conditions [128]. Rich media contained 2% dextrose (YPD). Synthetic complete media contained all amino acids except for those used for selection and 2% dextrose (SD, 2% Dex) or 2% galactose (2% Gal). Synthetic liquid media contained amino acids lacking the selective ones and 2% raffinose plus (SRGal) or minus 2% Gal (SRaf).

Induction of [PSI⁺] de novo

Yeast cells with Sup35 overexpression plasmids were grown in synthetic liquid selection media (SRaf) overnight. Unless otherwise stated, 2% Gal was added to the culture (OD∼0.5) to induce de novo Sup35 aggregation. In time course experiments where time exceeded 48 h, cultures were diluted back to OD∼0.5 in a fresh growth media to keep cells in exponential phase.

Staining of mature [PSI⁺] dots

After de novo [PSI⁺] was induced in [psi^-][PIN⁺] cells, they were grown on synthetic dropout media with 2% dextrose (SD) for many generations to maintain [PSI⁺] (∼8 days). Then, they were grown in 0.05% Gal for 3-4 h to allow Sup35NM-YFP to decorate existing [PSI⁺] aggregates.

Color and plate assay for [PSI⁺] induction

[PSI⁺] induced de novo was scored in yeast as described previously [33], [49], [72], [129], [130]. [PSI⁺] but not [psi^-] causes read-through of the nonsense mutation, ade1-14. The ade1-14 mutation causes the accumulation of a red by-product in the adenine biosynthesis pathway, so [psi^-] ade1-14 cells are red. In [PSI⁺] cells, when nonsense mutations are suppressed, ade1-14 cells become pink or white in rich media (YPD). In addition, the read-through of the ade1-14 allele in [PSI⁺] allows them to grow in media lacking adenine (SD-Ade), while [psi^-] cells cannot grow on this medium [49]. To confirm that Ade⁺ cells are of [PSI⁺] rather than suppressor mutants, they were mated with a tester strain ([psi^-] SUP35-GFP, GF658) to look for multiple fluorescent dots formed by endogenous Sup35-GFP in only [PSI⁺] cells.

FM4-64 and Thioflavin T staining

To locate the vacuole, cells were stained with FM4-64 as described previously [90]. Yeast cells were stained with Thioflavin T according to a protocol adapted from ref. [131] with the addition of two extra washes in PMST [0.1M KPO₄ (pH 7.5), 1 mM MgCl₂, 1 M Sorbitol, 0.1% Tween 20].

Visualization of aggregates and colocalization studies

Aggregates formed in cells by fluorescently labeled proteins were examined with a Nikon Eclipse E600 fluorescent microscope (100X oil immersion) and/or an Olympus FV1000 confocal microscope (60X oil immersion, with 1.6 magnifier). Colocalization was visualized by the confocal microscope using the channels of interest and by moving the focal plane up and down. Z-stacks were analyzed to confirm colocalization of proteins with 8-12 layers, with 0.5-1 µm increments. In dual color RFP/GFP studies, the RFP channel was always examined first to prevent visualization of activated GFP in the RFP channel.

Detection of protein levels

Upon the induction of overexpression of proteins tagged with VN or VC with galactose, cells were lysed as described previously [132]. Equal amounts of total proteins in precleared lysates were analyzed by Western blot using previously described antibodies [27], [63], [66], [104].

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12

Attachment 13

Attachment 14

Attachment 15

Attachment 16

Attachment 17

Attachment 18

Attachment 19

Attachment 20

Attachment 21