Heterologous Gln/Asn-Rich Proteins Impede the Propagation of Yeast Prions by Altering Chaperone Availability

Prions are self-propagating conformations of proteins that can cause heritable phenotypic traits. Most yeast prions contain glutamine (Q)/asparagine (N)-rich domains that facilitate the accumulation of the protein into amyloid-like aggregates. Efficient transmission of these infectious aggregates to daughter cells requires that chaperones, including Hsp104 and Sis1, continually sever the aggregates into smaller “seeds.” We previously identified 11 proteins with Q/N-rich domains that, when overproduced, facilitate the de novo aggregation of the Sup35 protein into the [PSI +] prion state. Here, we show that overexpression of many of the same 11 Q/N-rich proteins can also destabilize pre-existing [PSI +] or [URE3] prions. We explore in detail the events leading to the loss (curing) of [PSI+] by the overexpression of one of these proteins, the Q/N-rich domain of Pin4, which causes Sup35 aggregates to increase in size and decrease in transmissibility to daughter cells. We show that the Pin4 Q/N-rich domain sequesters Hsp104 and Sis1 chaperones away from the diffuse cytoplasmic pool. Thus, a mechanism by which heterologous Q/N-rich proteins impair prion propagation appears to be the loss of cytoplasmic Hsp104 and Sis1 available to sever [PSI +].


Introduction
The infectivity of transmissible spongiform encephalopathies (TSEs) was explained by the prion hypothesis proposing that the inheritance of biological information can be achieved by selfpropagating conformational changes in the prion protein PrP [1]. The prion list has since been extended to include protein-based genetic elements found in fungi [2]. The best-studied yeast prions [PSI + ], [PIN + ] (often called [RNQ + ]) and [URE3] are, respectively, self-propagating conformations of: Sup35, a translation termination factor; Rnq1, a protein of unknown function; and Ure2, a nitrogen catabolism repression regulator [3][4][5] [6,7]. The propagation of most [8][9][10], but not all [6] yeast prions is driven by their Q/N-rich prion domains that have the propensity to form aggregates in vivo and assemble into selfseeding, b-sheet-rich amyloid fibers in vitro [11,12].
Prion propagation involves templated conversion of soluble protein into the prion state [13] . In vitro data show that amyloid fibers grow by recruiting protein monomers to fiber ends [14]. In addition, prion propagation requires fibers to be fragmented to create new ends for conversion and to allow efficient transmission of seeds to daughter cells [15]. Failure at any of these steps would lead to loss (curing) of the prion.
When aggregated in the [PSI + ] state, Sup35's participation in translation termination is greatly reduced. This leads to increased read-through of stop codons, including the ade1-14 nonsense allele that can be readily monitored by a red/white color assay [16,[43][44][45]. [PSI + ] prion variants or strains manifest a range of distinct prion conformations that differ in levels of Sup35 aggregation and, consequently, in the frequency of stop-codon read-through. Since weak [PSI + ] variants cause less read-through of stop codons, are less mitotically stable and contain more of the soluble non-prion form of Sup35 than strong [PSI + ] variants, the color assay allows their distinction by the degree of red pigment accumulated [44,[46][47][48][49]. Also, these [PSI + ] variants differ in the size of their SDS-resistant Sup35 polymers [20].
When full length Sup35, or just its Q/N-rich prion domain, is transiently overproduced in [psi 2 ] cells, [PSI + ] is induced to appear, presumably because the excess protein increases the chance that it will form a prion seed [2,44,50,51]. However, efficient de novo induction of [PSI + ] requires the presence of a heterologous prion, e.g. [PIN + ] [3,51,52]. Heritable variants of the [PIN + ] prion have been distinguished by their efficiency in inducing [PSI + ] with the inducing efficiency gradually decreasing from very high to high to medium to low [PIN + ] [53].
In a screen of a high-copy yeast library for genes that enhance [PSI + ] induction in the absence of [PIN + ], an excess of any of 11 Q/N-rich proteins was found to promote de novo [PSI + ] appearance upon the overexpression of the prion domain of Sup35 [3]. Similarly, aggregation-prone polyQ sequences could substitute for [PIN + ] in the case of [PSI + ] induction, and [PIN + ] also facilitated the aggregation of proteins with extended polyQ stretches [54][55][56]. It was proposed that the aggregates formed by these Q/Nrich proteins provide a nidus for the formation of the first [PSI + ] seeds, which then promote Sup35's rapid aggregation. This crossseeding model postulates a direct interaction between Q/N-rich domains of a newly forming prion and preexisting heterologous prion or prion-like aggregates [3,54,57,58].
Several studies have indicated that heterologous prions or prion proteins can also inhibit prion propagation. For example, some [PIN + ] variants impede the inheritance of [PSI + ] [32,59] and [PSI + ] and [URE3] slightly destabilize each other [33,53]. Also, overexpression of the Ure2 prion domain or several other fragments of Ure2 cures [URE3] [60], and overexpression of some Rnq1 fragments encompassing the Q/N-rich C-terminal domain but lacking the N-terminus is inhibitory to [PSI + ] and [URE3] propagation in the presence of [PIN + ] [61]. Finally, overexpression of rnq1 N-terminal mutants causes enlargement of [PSI + ] aggregates leading to loss of [PSI + ] [62]. The molecular basis of these antagonistic interactions is unknown.
Here we report that overexpression of a number of Q/N-rich proteins can impede the propagation of the Q/N-rich prions, [PSI + ] and [URE3]. Our studies reveal a physical interaction between two such heterologous Q/N-rich protein aggregates and Hsp104. This hinders the availability of Hsp104 to shear prion aggregates, thereby inhibiting prion propagation. In contrast another overexpressed Q/N-rich protein does not sequester Hsp104, but rather appears to cure [PSI + ] by increasing the level of Hsp104.

Results
Overexpression of some Q/N rich proteins that eliminate the [PIN + ] requirement for the induction of [PSI + ] also destabilize pre-existing prions In an unsaturated genetic screen for overexpressed proteins that cure cells of [PSI + ], the most efficient curing was observed in the presence of a plasmid encoding a Q/N-rich portion of the CYC8 gene. Strikingly, CYC8 was one of the 11 genes we previously uncovered in a screen for genes that in high copy substitute for the [PIN + ] requirement for the de novo induction of [PSI + ] [3]. Therefore we asked if overproduction of the other proteins identified in the [PSI + ] induction screen would also destabilize preexisting [PSI + ]. Of the 11 chromosomal DNA fragments, 8 (STE18, YCK1, PIN2, URE2, PIN3, NEW1, NUP116 and LSM4) encode full-length proteins with Q/N-rich domains, and another 3 encode partial genes: the C-terminal Q/N-rich domains of PIN4 and CYC8, and the N-terminal Q/N-rich domain of SWI1, respectively, called here PIN4C, CYC8C and SWI1N.
Weak (w) [PSI + ][PIN + ] was transformed with the 11 multicopy plasmids with the URA3 and leu2-d markers [63], and encoding the above mentioned Q/N-rich proteins and protein fragments. Transformants grown on leucineless media that amplified the plasmids to a high-copy number were then examined for the presence of [PSI + ] using the color assay. This assay is based on the accumulation of a red pigment in ade1 mutants and the requirement of the Sup35 protein for proper termination at stop codons: cells in which much of the Sup35 release factor is sequestered into [PSI + ] aggregates are unable to efficiently terminate translation at the premature stop codon in ade1-14, and some full-length Ade1 is synthesized despite the mutation. Thus, ade1-14 cells that give rise to white or pink colonies are [PSI + ], while those that grow into red colonies are [psi 2 ].

Author Summary
Certain proteins can occasionally misfold into infectious aggregates called prions. Once formed, these aggregates grow by attracting the soluble form of that protein to join them. The presence of these aggregates can cause profound effects on cells and, in humans, can cause diseases such as transmissible spongiform encephalopathies (TSEs). In yeast, the aggregates are efficiently transmitted to daughter cells because they are cut into small pieces by molecular scissors (chaperones). Here we show that heritable prion aggregates are frequently lost when we overproduce certain other proteins with curing activity. We analyzed one such protein in detail and found that when it is overproduced it forms aggregates that sequester chaperones. This sequestration appears to block the ability of the chaperones to cut the prion aggregates. The result is that the prions get too large to be transmitted to daughter cells. Such sequestration of molecular scissors provides a potential approach to thwart the propagation of disease-causing infectious protein aggregates.
[URE3] is also destabilized by high copy plasmids encoding Q/N-rich domains Changes in the state of strong [PSI + ] caused by excess Pin4C were monitored over time. After overnight overexpression of Pin4C, the Sup35-GFP foci became brighter, bigger and more distinct in ,80% of the cells in comparison with the numerous tiny Sup35-GFP foci formed in control cells lacking the Pin4C plasmid ( Figure 2A). Likewise, Pin4C overexpression caused the size distribution of SDS-resistant [PSI + ] polymers to shift dramatically to larger complexes ( Figure 2A). These enlarged Sup35-GFP foci that formed following overnight expression of Pin4C were still capable of propagating [PSI + ] if expression of Pin4C was turned off. However, when Pin4C was expressed for another 4 days, Sup35-GFP became diffuse and [PSI + ] was lost ( Figure S2).
Overexpressed Pin4C forms amyloid-like aggregates, which do not colocalize with [PSI + ] aggregates  Figure 2B). The formation of the large Pin4C-DsRed dot was always accompanied by the appearance of large Sup35-GFP foci within the cell, but they did not colocalize ( Figure 2B). Furthermore, although Pin4C aggregates have the SDS-resistant characteristic of amyloid aggregates ( Figure 2C), the sizes of Sup35 and Pin4C SDS-resistant polymers were not identical ( Figure 2C), suggesting that Pin4C is not a component of SDS-resistant [PSI + ] polymers.

PIN4C-induced increase of [PSI + ] aggregate size requires continuous synthesis of Sup35
The large Sup35 aggregates that appeared in the presence of excess Pin4C could have been formed by the simple association of existing [PSI + ] aggregates, or by enlargement of individual aggregates, e.g. due to reduced shearing of growing amyloid fibers. The first possibility is modeled on our previous finding that overexpressed Sup35 causes pre-existing [PSI + ] aggregates to coalesce into larger particles [68]. Thus we reasoned that overproduced Pin4C might ''glue'' existing [PSI + ] aggregates together through heterologous Q/N-rich domain interactions.
To test this, we examined the appearance of pre-existing [PSI + ] aggregates when Pin4C was overexpressed. Pre-existing aggregates were made of protein encoded by extrachromosomal SUP35-GFP controlled by the repressible TET r promoter in a strain lacking the endogenous SUP35 prion domain (SUP35DNM). After 6 hrs of expression of the GAL controlled PIN4C-DsRED (i.e. before diffuse  ] expressing SUP35DNM at its endogenous locus and harboring extrachromosomal pTET r -SUP35-GFP (L3126) were transformed with pHR81GAL-PIN4C-DsRED, or vector, pHR81GAL-DsRED. Cells were grown in 2% raffinose +2% galactose +0.025 mg/ml doxcycline for 6 hrs, which induced PIN4C-DsRED and allowed the Sup35-GFP level to be close to the normal Sup35 level. Single cells with diffuse DsRed were then micromanipulated and grown for 18 hrs dividing ,3 times on 2% raffinose +2% galactose +10 mg/ml doxcycline medium where new synthesis of Sup35-GFP is repressed. GFP and DsRed images of a representative part of a growing microcolony are shown. The arrows point to mother cells, and the arrowheads point to two daughter cells with and without a large DsRed dot. doi:10.1371/journal.pgen.1003236.g002 Pin4-DsRed examined at the same time ( Figure 2D). The faint appearance of the GFP aggregates was due to lack of newly synthesized Sup35-GFP. The absence of large Sup35-GFP foci even in cells with large Pin4C-DsRed aggregates, suggests that the large Sup35-GFP foci seen in the presence of continued Sup35-GFP synthesis are not formed by coalescence of previously formed [PSI + ] aggregates. Rather, [PSI + ] aggregates appear to become larger by continuously incorporating newly synthesized Sup35-GFP upon overexpression of Pin4C.
Pin4C overexpression reduces Sup35-GFP aggregate mobility and transmission to daughter cells The dynamics of Sup35-GFP in dividing [PSI + ] cells upon Pin4C overexpression was probed using fluorescence recovery after photobleaching (FRAP). The rate of transfer of Sup35-GFP from mother to daughter cells was examined by first completely photobleaching daughter cells and then measuring the fluorescence recovery of the daughter ( Figure 3A). As shown previously [69,70], soluble Sup35-GFP in [psi 2 ] cells was much more mobile than predominantly aggregated Sup35 in [PSI + ] cells; the average half-time for recovery in isogenic [psi 2 ] versus [PSI + ] daughters was, respectively, 7 s versus 63 s ( Figure 3B, 3C and 3E).
The fluorescence recovery measured following Pin4C overexpression in 12 [PSI + ] cells containing large Sup35-GFP foci indicated that the population of Pin4C expressing cells is heterogeneous. Three cells exhibited almost no recovery, indicating a major defect in transmission of Sup35-GFP ( Figure 3D (I) and 3E). In another 4 cells, Sup35-GFP only recovered to 67% of the intensity observed prior to photobleaching, with a half-time of 126 s that is twice as long as in [PSI + ] without Pin4C overexpression ( Figure 3D (II) and 3E). In one cell 98% recovery was completed with a half-time of 243 s (data not shown). Yet, in 3 cells, fluorescence recovered to 100% with half-times similar to that in [PSI + ] cells without Pin4C ( Figure 3D (III) and 3E). We also observed one cell (data not shown) exhibiting 83% recovery with a half-time of only 17.79 s.
The slow flux of Sup35-GFP in 8 of these mother-daughter pairs indicates that Sup35-GFP often becomes extremely immobile following overexpression of Pin4C, which is consistent with an increase in Sup35 aggregate size and a reduction in the segregation of prion seeds to daughters. However, the presence of cells with normal recovery suggests that at least in some cells, in addition to the large Sup35-GFP foci, there were still small prion seeds available to be transmitted to daughters. Finally, the existence of cells with very fast flow of Sup35-GFP from mother to daughter cells indicates a high level of soluble Sup35 that might be already inefficiently sequestered by the few large Sup35 aggregates still remaining in the mother cell. Differences in the rates of Sup35-GFP transfer among individual mother-daughter pairs suggest that overexpressed Pin4C creates heterogeneity in the properties of prion aggregates during the curing process.

Microcolonies overexpressing Pin4C show progressive loss of [PSI + ]
To further assess how the appearance of large immobile Sup35 aggregates caused by excess Pin4C correlates with the loss of [PSI + ], individual cells of the [PSI + ] Sup35-GFP strain carrying pHR81GAL-PIN4C-DsRED were micromanipulated and grown on 2% raffinose + 2% galactose plates where the DsRed tagged Pin4C was expressed. As the cells divided we monitored the changes of Sup35-GFP distribution in the cells within the microcolonies. The outer edges of microcolonies with a single layer of cells were imaged since the central portion of the microcolony included multiple layers of overlapping cells. In the edge of one sector (Figure 4, upper panel), Sup35-GFP remained in the multiple tiny foci seen in [PSI + ] cells prior to Pin4C induction. But in the edge of another sector (Figure 4, lower panel), Sup35-GFP foci increased in size and were reduced in number progressively in dividing cells, which eventually segregated out [psi 2 ] cells (also see Figure S3). Different phenotypes observed in different sectors may be due to differences of PIN4C plasmid copy number within individual cells.

Cell division is required for overexpression of Pin4C to cure [PSI + ]
Previous studies showed that cell division was essential for the loss of the [PSI + ] prion in GuHCl-treated cells [71]. To test if cell division was required for the overexpression of Pin4C to cure [PSI + ], loss of [PSI + ] was compared in a MATa strong [PSI + ] SUP35-GFP strain overexpressing Pin4C in the presence and absence of growth arrest induced by a-factor. Pin4C was overexpressed in liquid galactose for 40 hrs, and 50 mM a-factor was added at this stage, i.e. when Sup35-GFP aggregates were larger and fewer in number, but before the emergence of any diffuse [psi 2 ] cells. After overexpressing Pin4C for another 16 hrs, cells were plated on YPD to score for [PSI + ] loss. The a-factor arrest caused an 88% decrease in colony-forming units, CFUs. Cultures whose growth was arrested vs. not arrested respectively showed, 1% vs. 59% loss of [PSI + ] ( Figure 5). Reduced loss of [PSI + ] during the a-factor arrest indicates that cell division is required for curing of [PSI + ] by overexpressed Pin4C.

Overexpression of Pin4C does not change chaperone levels
Since [PSI + ] propagation is sensitive to optimal levels of chaperones such as Hsp104, Ssa1/2, Ssb1, Sse1 and Sis1 [72][73][74], it seemed possible that excess Pin4C caused a change in levels of chaperones which then led to loss of [PSI + ]. However, Pin4C overexpression did not cause a significant alteration in levels of Ssa1/2, Ssb1, Sse1 and Sis1 ( Figure S4A).
More thorough analysis revealed that the level of Hsp104 in cells following Pin4C overexpression was reduced to 83% of that in cultures not overproducing Pin4C ( Figure S4B and Table S1). Because previous studies showed that a heterozygous disruption of HSP104 has no effect on [PSI + ] propagation [75], and that loss of [PSI + ] is only initiated when the Hsp104 levels drop well below 50% of the normal level [76], it appeared unlikely that the slight decrease in Hsp104 level induced by excess Pin4C would cause [PSI + ] loss. Indeed, a heterozygous disruption of HSP104 did not facilitate curing of [PSI + ] in our strains ( Figure S4C).

Overexpressed Pin4C titrates Hsp104-GFP away from the cytoplasm
The increased size of Sup35 polymers seen upon Pin4C overexpression was similar to that seen upon inhibition of Hsp104 due to a block of prion fragmentation [20]. Thus we considered the possibility that excess Pin4C cures [PSI + ] by titrating Hsp104 away. Since [PSI + ] aggregates were found to associate with Hsp104 [38], it seemed possible that Pin4C aggregates also harbor Hsp104.
To visualize the distribution of Hsp104, we used the Hsp104-GFP strain from the endogenously GFP-tagged yeast library [77]. In unstressed cells, Hsp104-GFP is observed as diffuse GFP or occasionally tiny foci with diffuse background ( Figure 6A). However, after 16 hrs of induction of untagged Pin4C, Hsp104-GFP coalesced into one large aggregate per cell. Such large Hsp104-GFP aggregates were never found in control cells without Pin4C overexpression. When Pin4C tagged with DsRed was used, the big Pin4C-DsRed focus colocalized with the coalesced Hsp104-GFP ( Figure 6A). Furthermore, Hsp104 was co-immuno-   [PSI + ]. Also, when expressed at the normal level, Hsp104 T160M maintains [PSI + ] [26]. Therefore we overexpressed the Hsp104 T160M mutant allele. Excess Hsp104 T160M did not reduce Pin4C aggregation (data not shown), however it reduced the efficiency with which overexpressed Pin4C caused loss of [PSI + ] ( Figure 6C).

Increased levels of Sis1 prevent aggregation of Pin4C and reduce the ability of Pin4C to cure [PSI + ]
Sis1 is a chaperone involved in cleaving [PSI + ] aggregates and generating new prion seeds. It was hypothesized to recruit Hsp104 to the sites of prion aggregation [39]. We observed that overexpressed Pin4C also sequestered Sis1. As described previously [78], without Pin4C overexpression most Sis1-GFP was found in the nucleus ( Figure 8A). Upon Pin4C-DsRed overexpression, much of the Sis1-GFP colocalized with cytoplasmic Pin4C-DsRed aggregates, and the amount of Sis1-GFP remaining in the nucleus was significantly reduced (Figure 8). Thus we asked if increased levels of Sis1 would affect curing of [PSI + ] by overexpression of Pin4C. Indeed, we found that the loss of [PSI + ] by overexpression of Pin4C was significantly reduced by overexpression of Sis1 ( Figure 9A). Furthermore, in cells with excess Sis1, overproduced Pin4-DsRed accumulated in several small foci and did not form the huge single focus observed in the absence of excess Sis1. Also, in cells with excess Sis1, Sup35-GFP remained in multiple tiny foci, that did not enlarge upon Pin4C overexpression ( Figure 9B). Thus, overproduced Sis1 prevents overexpressed Pin4C from forming big foci and reduces the formation of large Sup35 aggregates, which may cause decreased [PSI + ] curing by Pin4C.

Effects of overexpressed Pin3 and Cyc8C on chaperones
To investigate if titrating Hsp104 is a general mechanism by which the heterologous Q/N-rich proteins cure prions, we visualized Hsp104-GFP in cells overexpressing Pin3 or Cyc8C. Like Pin4C, overexpressed Pin3 caused Hsp104-GFP to coalesce into large aggregates and reduced the level of diffuse cytoplasmic Hsp104-GFP fluorescence relative to empty vector controls. However, overexpressed Pin3 also caused a slight increase in the cellular levels of Hsp104, Sse1, Ydj1 and Sis1 ( Figure S6A). The combined result of these two effects was that the Hsp104 cytoplasmic level was about 74% of that seen in cells without Pin4C overexpression ( Figure 7C). Likewise, although overexpressed Pin3 caused Sis1-GFP to coalesce (Figure 8A), the levels of Sis1 that remained in the cytoplasm were similar to controls ( Figure 8B). In contrast to overexpressed Pin4C or Pin3, overexpressed Cyc8C caused an 8-fold increase in the Hsp104 level ( Figure S6B), and did not sequester Hsp104 ( Figure 7B). Overexpressed Cyc8C also caused a slight increase in the Sis1 level ( Figure S6B).

Overexpressed Gpg1 may titrate chaperones away from the cytoplasm
We next investigate if non-Q/N-rich aggregates that cure [PSI + ] also sequester chaperones. Gpg1 is a mimic of a G protein c subunit. Like overexpressed Pin4C, overexpressed Gpg1 formed aggregates, had reduced curing efficiency when Hsp104 was overexpressed, but did not affect the cellular levels of Hsp104 [79]. Previous work visualizing Gpg1 curing of [PSI + ] aggregates was complicated by the use of overexpressed Sup35NM-GFP [79]. When we examined the effect of excess Gpg1 on fluorescent aggregates in [PSI + ] cells with endogenous Sup35 tagged with GFP, we found that the fluorescent dots got larger and fewer in number ( Figure 10A), just as seen when Pin4C was overexpressed. Furthermore, like excess Pin4C, excess Gpg1 caused endogenous Hsp104 tagged with GFP to aggregate into foci. However, despite this aggregation we did not detect any reduction in the intensity of cytoplasmic Hsp104-GFP compared to the vector control ( Figure 10B).

Discussion
Surprisingly, several factors that enhance prion induction also cause prion destabilization: [ [80]. We now report that overexpression of many of the 11 Q/N-rich proteins, which in high copy substitute for the [PIN + ] requirement for [PSI + ] induction, also destabilize pre-existing [PSI + ] or [URE3] (Figure 1), indicating that Q/N-rich proteins can also both enhance prion appearance and impair propagation of existing prions.
Our finding that the high copy plasmids encoding 11 Q/N-rich domains can promote the de novo induction of [PSI + ] in an rnq1D strain, establishes that the induction of [PSI + ] does not first require the appearance of [PIN + ]. Rather, this suggests a direct interaction between heterologous Q/N-rich proteins and Sup35, and, as originally hypothesized [3], and as indicated by earlier in vitro studies [54,58], that the Q/N-rich proteins themselves are likely substituting for [PIN + ] as the cross-seeds.
Since Other previously hypothesized mechanisms of prion curing by Q/N-rich proteins are that the overexpressed Q/N-rich domain of Ure2 inhibits prion fiber growth leading to curing of [URE3], either by incorporating into the growing tip of the [URE3] seed thereby blocking or ''capping'' its growth [60,81,82], or by sequestering Ure2 preventing it from joining prion fibers [81,82]. However, this mechanism implies a very efficient interaction between the curing protein and the prion-forming protein, which is likely only in the case of very high homology.
Our studies indicate that overexpression of Pin4C allows [PSI + ] aggregates to continuously incorporate soluble Sup35, but prevents proper fragmentation ( Figure 2, Figure 3, Figure 4, Figure 5). Thus, overproduced Pin4C causes [PSI + ] loss via a defect in breakage of growing [PSI + ] fibers and transmission of prion seeds. This is quite different from either the capping or sequestration of its own protein models. Since prion fragmentation is crucially dependent on chaperones, particularly on Hsp104, overexpressed Pin4C could affect the function of Hsp104 on [PSI + ]. One possibility is that Pin4C coats [PSI + ] aggregates through the interaction of their Q/N-rich domains and thus shields Sup35 polymers from the shearing activity of chaperones. However, we did not detect co-localization of Sup35-GFP and Pin4C-DsRed foci ( Figure 2B).
Although there is no significant alteration in total levels of chaperones ( Figure S4), excess Pin4C sequesters Hsp104 and Sis1 ( Figure 6, Figure 7, Figure 8), reducing their availability to shear [PSI + ] aggregates. Although depletion of Sis1 only causes delayed and gradual loss of [PSI + ] [35], the reduced availability of Sis1 may enhance the effect of Hsp104 sequestration. Furthermore, although excess Sis1 prevented curing of [PSI + ] by Pin4C overexpression, excess Sis1 also decreases Pin4C aggregation (Figure 9).
We also observed that different Q/N-rich proteins had distinct effects toward the yeast Q/N-rich prions, [PSI + ] and [URE3] (Figure 1). For example, our data is consistent with the recent study showing that overproduction of New1 does not cure [PSI + ] [83], but we further uncover that overexpressed New1 does cause efficient curing of [URE3]. Likewise, Pin3 cures [URE3] but not [PSI + ]. One explanation could be that these proteins sequester Hsp104 less efficiently than the proteins that cure both prions such as Pin4C. Indeed, we found that overexpressed Pin3 reduced the cytoplasmic level of Hsp104 less effectively than Pin4C (Figure 7 and Figure S6A). Since more than 50% of the cellular Hsp104 Pin4C-DsRed complex immunocaptured with anti-DsRed from 500 mg of total protein was loaded as ''eluate''. The same membrane was immunoblotted (IB) with anti-DsRed, then with anti-Hsp104, and re-probed with anti-Pgk1 as a control. No co-immunocapture of endogenous Pgk1 with Pin4C was detected, implying that Hsp104 was specifically immunocaptured with the Pin4C complex. The slightly slower migration of Hsp104 in the ''eluate'' relative to its migration in the ''lysate'' is probably due to the different buffers used during immunocapture. (C) Overexpression of Hsp104 T160M suppresses curing of strong [PSI + ] by Pin4C. Transformants with pHR81GAL-PIN4C and pRS413GAL-HSP104 T160M (qPin4C, qHsp104 T160M ); or with pHR81GAL-PIN4C and empty vector pRS413GAL (qPin4C); or with pRS413GAL-HSP104 T160M and pHR81GAL (qHsp104 T160M ); or with both empty vectors pHR81GAL and pRS413GAL (qvectors) were selected on plasmid selective glucose medium, replica-plated to plasmid selective inducing galactose medium, and then 10-fold serially diluted (10 5 R10 0 cells from left to right) and spotted onto glucose YPD medium where expression of Pin4C and Hsp104 T160M is turned off. There was no growth inhibition in cells overexpressing Pin4C and Hsp104 T160M compared to those overexpressing Pin4C alone when spotted on a galactose plate ( Figure S5) remained diffuse when Pin3 was overexpressed, curing of [PSI + ] would not be expected [76]. However, since there are fewer prion seeds in [URE3] than weak [PSI + ] cells [35], this slight reduction in soluble Hsp104 might be sufficient to cure [URE3]. Indeed, previous studies showed that different prions and different prion variants have different susceptibilities towards chaperone activities [37,69,84]. Although overexpressed Pin3 sequestered Sis1 ( Figure 8A), there was no significant difference in the fluorescence intensity of diffuse Sis1-GFP in the cytoplasm with Pin3 overexpression compared to the control without Pin3 overexpression ( Figure 8B), suggesting that sequestration of Sis1 by excess Pin3 does not significantly contribute to curing of [URE3]. There was also a slight increase in the total cellular levels of Sse1 and Ydj1, which could contribute to the Pin3 curing.
Surprisingly, overexpression of Cyc8C cures [PSI + ] but not [URE3] (Figure 1). This cannot be explained by sequestration of Hsp104. Indeed, overexpression of Cyc8C did not cause aggregation of Hsp104 ( Figure 7B), but rather increased the Hsp104 level 8-fold while having no effect on Ssa1 ( Figure S6B). Since overexpression of Hsp104 cures [PSI + ] but not [URE3] [16,42], this provides a plausible mechanism. We also observed a slight increase in the Sis1 level by overexpressed Cyc8C ( Figure  S6B). It was previously shown that overproduced Sis1 enhances curing of [PSI + ] by overexpressed Hsp104 [80], therefore overexpressed Cyc8C may cure [PSI + ] through additive effects of the increased level of Hsp104 and Sis1.
Our findings may provide an explanation for previous observations that overexpressed proteins lead to curing of prions. Overexpression of the Rnq1D100 protein (the Q/N-rich Cterminal domain) eliminates [PSI + ] and [URE3] in the presence of [PIN + ] [61]. Also mutations in the non-Q/N-rich domains of RNQ1 cause an increase in the size of Sup35 aggregates leading to curing [62]. Both of these phenomena could be because the Rnq1 fragment or mutants form [PIN + ]-dependent aggregates that sequester Hsp104 and/or other chaperones and reduce their availability to aid [PSI + ] propagation.
The chaperone titration curing mechanism may also be applicable to non-Q/N-rich aggregates. Indeed, a similar mechanism could explain prion curing caused by overexpression of the non-Q/N-rich protein Gpg1 (Figure 10). Although Gpg1 sequestered Hsp104, we could not detect a reduction in cytoplasmic Hsp104. Nonetheless, other chaperones might also be sequestered by Gpg1 aggregates leading to reduced prion shearing and prion loss.
A recent study also indicates that the cellular localization of chaperones can have a direct impact prion propagation. Indeed, this appears to explain the long-standing mystery of why overexpression of Hsp104 cures cells of [PSI + ] but not of other prions [16,36,42,51]. Hsp104 overexpression was shown to inhibit shearing of [PSI + ] aggregates because the excess of Hsp104 displaced the Hsp70 chaperone Ssa1 from the [PSI + ] aggregate. Hsp104 does not bind to other prions in the absence of Ssa1, consistent with the absence of curing [29].
Other studies also indicate that titration of cellular proteins by amyloid aggregates can have a profound effect on the cell. Indeed, sequestration of essential proteins by amyloid aggregates was previously shown to cause prion toxicity. The large Sup35 aggregate that forms when Sup35 is overexpressed in [PSI + ] cells sequesters the essential Sup35 binding partner Sup45, resulting in death [68]. Likewise, the large Rnq1 aggregates that form when Rnq1 is overexpressed in [PIN + ] cells sequester the core spindle pole body component Spc42 causing toxicity [85]. Also, polyglutamine (polyQ) aggregates sequester essential endocytic components such as Sla2 [86] and endoplasmic reticulum associated degradation proteins in [PIN + ] dependent toxicity [87], and sequester Sup35 in [PSI + ]-dependent polyQ toxicity [88].
Our results establish sequestration of specific chaperones by overexpressed proteins as a general mechanism to alter the cellular localization of chaperones and therefore inhibit prion propagation. Similar mechanisms could influence phenotypic variation by regulating the balance of chaperones needed for prion propagation, in response to environmental stimuli. Since biochemical pathways controlling prion formation and/or maintenance appear to be conserved from yeast to mammals, titration of chaperones via heterologous Q/N-rich aggregates might provide a new approach to prion and amyloid disease intervention.

Strains, media, and plasmids
All strains used are described in Table 1 ] cytoductants were confirmed by their inability to grow on medium lacking histidine. GF708 transformed with pURE2-URE2N::GFP-HIS3 was crossed to strains to score for their [URE3] state which was indicated by the appearance of fluorescent foci.
Standard yeast media were used [92]. For overexpression of library high copy plasmids [93] transformants were selected on plasmid selective synthetic media with dextrose lacking uracil (SD-Ura) and then spread on synthetic media lacking both uracil and leucine (SD-Ura-Leu) to amplify the copy number of leu2-d bearing plasmids about 100 fold [63]. For Pin4C overexpression from pHR81GAL-PIN4C in liquid medium, cultures were grown in 2% raffinose synthetic media lacking uracil for ,8 hrs and then transferred to 2% raffinose + 2% galactose media lacking uracil and leucine (SGal-Ura-Leu) for ,16 hrs. Transformants carrying double plasmids were selected on SD-Ura-Trp and replicated to SGal-Ura-Trp-Leu to induce overexpression of the GAL controlled genes on both plasmids. The high copy genomic library used in our earlier study was constructed in the pHR81 (2 m URA3 leu2-d) vector [3,93]. The PIN4C ClaI-XhoII fragment isolated from the PIN4 library clone #277 [3] was Klenow-filled and cloned into pHR81H (2 m HIS3 leu2-d, with the URA3 marker of pHR81 exchanged for the HIS3 marker) vector at the blunt-ended BamHI site to generate pHR81H-PIN4C (2 m HIS3 leu2-d). The GAL promoter isolated from pRS316-GAL1 (a kind gift from A. Bretscher) on the XhoI fragment was filled in and cloned into the unique pHR81 BamHI site to create pHR81GAL. The PIN4C fragment isolated from the PIN4 library clone #277 [3] as an XhoII fragment was filled in and cloned into pHR81GAL at the BamHI site to generate pHR81GAL-PIN4C. DsRED as a NdeI-NotI fragment digested from pDsRed-Monomer-N1 vector (Clontech) was filled in and cloned into pHR81GAL at the BamHI site to produce pHR81GAL-DsRED. The PIN4C fragment was PCR amplified using primers with BamHI linkers and subcloned into pHR81GAL-DsRED at the BamHI site to produce pHR81GAL-PIN4C-DsRED. The PCR primers were P1 (59-ggctcgagtggatcggcgggaggaaattgaaag-39) and P2 (59agtggatcctcgagtggtacctctagaagtatataataccatagattc-39). The SUP35-GFP fragment isolated as a SacI-BamHI fragment from p1744 (SB238, kindly provided by T. R. Serio) was subcloned into pTETr vector p1331 (a kind gift of pCM184 from E. Herrero) to create pTETr-SUP35-GFP. Plasmid pGAL-HSP104 T160M is a kind gift from D. C. Masison [26]. Plasmid pYES3GAL-SIS1 is a kind gift from S. L. Lindquist [94]. GPG1 isolated from pGPD-GPG1 (a kind gift from Y. Nakamura and H. Kurahashi) [79] on the BamHI-XhoI fragment was filled in and cloned into the unique pRS316-GAL1 BamHI site to create pRS316-GAL1-GPG1.

Scoring for loss of prions
A weak [PSI + ][PIN + ] variant of 74-D694 (L1758) was transformed with the pHR81 based high copy vectors with inserts encoding any of 9 Q/N-rich domains: PIN4C clone #277, CYC8C clone #151, STE18 clone #299, YCK1 clone #103, PIN2 clone #222, URE2 clone #155, PIN3 clone #80, NEW1 clone #39 and LSM4 clone #288, whose overexpression substitutes for [PIN + ] in [PSI + ] induction [3]. For each clone, three transformants selected on SD-Ura were spread on SD-Ura-Leu where the library plasmids were amplified to high copy number because of their poorly expressed leu2-d allele [63]. Three equal size colonies were resuspended in water and spread on YPD where the percentage of red [psi 2 ] colonies was scored. A similar method was used to score for loss of strong [ ] (L1953) strains were transformed with pHR81 empty vector or pHR81-PIN4C. Two transformants selected on SD-Ura were spread on SD-Ura-Leu to overexpress Pin4C. Three equal size colonies of each transformant were resuspended in water and spread on YPD. Ten colonies from each clone were crossed to [pin 2 ] 64-D697 harboring pCUP1-RNQ1::GFP-TRP1 and diffused Rnq1-GFP was scored as [pin 2 ].

Fluorescence microscopy
Yeast expressing Sup35-GFP and Pin4C-DsRed were imaged with a Zeiss Axioskop 2 microscope. For time lapse experiments, single cells were micromanipulated onto a 2% agar patch, then covered with a coverslip and placed on 2% raffinose + 2% galactose plates to allow cell growth. The patch with the coverslip in place was then transferred to a glass slide to image the microcolony and was returned to the plate for further growth. For colocalization experiments, yeast expressing Hsp104-GFP were grown overnight in 2% glucose and then induced in 2% raffinose + 2% galactose liquid medium to overexpress Pin4-DsRed.

FRAP
Fluorescence recovery after photobleaching (FRAP) was performed on a Zeiss LSM510 Axiovert confocal microscope. Mothers with buds smaller than 2.5 mm in diameter were selected in GF657 where Sup35-GFP foci became larger following Pin4 overexpression. Buds were completely bleached with a 488-nm laser at 100% power. After photobleaching, single scan images were collected every 5s with 3% laser and 56 zoom power. The pinhole was fully open to allow complete bleaching and to yield enough signal for fluorescent recovery.
Relative fluorescence intensity (RFI) was determined by RFI = ((Ne t -Ne min /N1 t )/(Ne 0 -Ne min /N1 0 ))6100, where Ne t is the average intensity of the bleached bud at time t and N1 t is the average intensity of its non-bleached mother cell at the corresponding time used to compensate for loss in total fluorescence [95]. Ne 0 and N1 0 represent average intensities of the bleached bud and its non-bleached mother respectively before photobleaching. Ne min is the minimum fluorescence intensity of the bud seen. The half-time that indicates the speed of mobility and the plateau level of recovery were measured by curve fitting the RFI data to a one-phase exponential association alogorithm with GraphPad Prism.

Quantification of fluorescence intensity
Fluorescent images were acquired using the same exposure time for all the samples. Using ImageJ, cytoplasmic regions devoid of aggregates and the vacuole were selected with the ''brush'' selection tool. The mean fluorescence intensity in the selected area was quantified using the ''measure'' function. The area around the cell was selected as the background. The data for each cell was obtained by calculating the mean fluorescence in the cytoplasm subtracted by that in the background.

a-factor arrest
Cell growth was arrested by the addition of 50 mM of the yeast mating pheromone a-factor. a-factor peptide (Trp-His-Trp-Leu-Gln-Leu-Lys-Pro-Gly-Gln-Pro-Met-Tyr) was from GenScript. Preparation and analysis of yeast cell lysates Cells overexpressing PIN4 or PIN4-DsRED were grown in 150 ml of 2% raffinose +2% galactose media to an A 600 OD of 0.3-0.8, where 80% of cells contained larger Sup35-GFP foci. Lysates were prepared as described [32]. For chaperone analysis, equal amounts of total proteins in precleard lysates were analyzed by Western blot using previously described antibodies [38]. Monoclonal anti-Pgk1 antibody was from Invitrogen. For semidenaturing detergent agarose gel electrophoresis (SDD-AGE), 50-80 mg of total protein in precleared lysates were incubated for 7 min in sample buffer with 2% SDS at room temperature and resolved on 1.5% agarose gels [96].

Immunocapture of cell lysates on magnetic beads
Immunocapture experiments were essentially as described [38] with the following changes: 750 ml of a higher salt lysis buffer [LB2: 40 mM Tris-HCl (pH 7.5), 150 mM KCl, 5 mM MgCl 2 , 10% glycerol] was used; 500 ml lysates of 0.5-1.0 mg/ml proteins were incubated with 3 ml of a-DsRed antibody for 2 hrs on ice; samples were mixed with 50 ml magnetic beads with immobilized G protein (Miltenyi Biotec) and incubated on ice for 30 min. Finally, beads were washed with 1.0 ml of each of the following solutions at 4uC in the following order to remove nonspecifically bound proteins: LB2 with 1% Triton X-100; LB2, 210 mM KCl, 1% Triton X-100; LB2 with 1% Triton X-100; LB2; LB1 [40 mM Tris-HCl (pH 7.5), 50 mM KCl, 5 mM MgCl 2 , 5% glycerol]; 20 mM Tris-HCl (pH 7.6). a-DsRed for immunocapturing was monoclonal antibody from Clontech, and a-DsRed for detection was a polyclonal antibody from Santa Cruz Biotechnology. Figure S1 There is no growth advantage of [psi 2 ] over [PSI + ] cells upon overexpression of Q/N-rich domains. Amplification of high copy plasmids encoding Q/N-rich domains exhibits no differences in cell viability in the absence or presence of [PSI + ]. Isogenic [PIN + ] strains lacking (2) [PSI + ] (L1749) or containing (+) weak [PSI + ] (L1758) were each transformed with the high copy (URA3, leu2-d) plasmids with an insert encoding the indicated Q/ N-rich domain, or the control empty vector pHR81. Two transformants for each plasmid were 10-fold serially diluted and spotted on SD-Leu to amplify library plasmids (left panels), and SD-Ura to maintain plasmid low copy number (right panels). One representative transformant for each plasmid was photographed after 4 days of incubation.  (GF657) cells carrying pHR81GAL-PIN4C-DsRED were micromanipulated and grown on 2% raffinose + 2% galactose to induce Pin4C-DsRed for ,24 hrs. A portion of the microcolony is shown as a GFP image. Sup35-GFP foci increased in size and were reduced in number progressively in cells dividing from the center to the edge of the microcolony. Single huge faint fluorescent areas in some cells are due to leakage of Pin4C-Dsred foci into the GFP channel; such foci were never observed in the GFP channel when overexpressing the Pin4C not tagged with DsRed (see Figure 2A).

Supporting Information
(TIF) Figure S4 The Hsp104 level is slightly reduced following Pin4C overexpression. (A) Overexpression of Pin4C has no notable effect on the expression levels of chaperones. Lysates of strong [PSI + ][PIN + ] (GF657) following overnight overexpression of Pin4C from pHR81GAL-PIN4C were analyzed by immunoblotting with the indicated antibodies. Ribosomal protein L3 was used as an internal loading control. Control cultures were transformed with the pHR81GAL vector. (B) Hsp104 expression was visualized using a PhosphorImager scanning system after immunoblotting the lysates described above with anti-Hsp104 antibody and also with anti-Pgk1 antibody. (C) A heterozygous disruption of HSP104 has no effect on [PSI + ] propagation. Genomic SUP35-GFP strong [PSI + ][PIN + ] (GF845) carrying pHR81GAL-PIN4C (qPin4C, HSP104), or diploids from a cross of GF845 harboring the empty vector pHR81GAL to a [psi 2 ] strain with a disruption of HSP104 and genomic SUP35-GFP (GF844) harboring the empty vector pRS413 (qvectors, HSP104/D), were grown on plasmid selective glucose medium, and replica-plated onto plasmid selective galactose to induce the GAL promoter, and then 10-fold serial diluted (10 5 ,10 0 cells from left to right) and spotted onto YPD glucose medium. Shown is a representative image. There were no red colonies indicative of [psi 2 ] observed in the HSP104 heterozygous disruption background. (TIF) Figure S5 There is no difference in cell growth with overexpressed Pin4C in the presence and absence of excess Hsp104 T160M . Strong [PSI + ][PIN + ] SUP35-GFP cells (GF657) with pHR81GAL-PIN4C and pRS413GAL-HSP104 T160M (qPin4C, qHsp104 T160M ); or with pHR81GAL-PIN4C and empty vector pRS413GAL (qPin4C); or with pRS413GAL-HSP104 T160M and pHR81GAL (qHsp104 T160M ); or with both empty vectors pHR81GAL and pRS413GAL (qvectors) were grown on plasmid selective glucose medium, and then 10-fold serially diluted (10 4 ,10 2 cells from left to right) and spotted onto plasmid selective galactose to induce the GAL promoter. Transformants spotted onto plasmid selective glucose medium (-Ura-Trp) were used as a control. (TIF) Figure S6 Effects of overexpression of Cyc8C or Pin3 on chaperone levels. (A) Overexpression of Pin3 caused a slight increase in the levels of Hsp104, Sse1, Sis1 and Ydj1. Lysates of cells with GFP tagged endogenous Hsp104 following overnight overexpression of Pin3 from pHR81GAL-PIN3 were analyzed by stripping and immunostaining the same blot with the indicated antibodies, except that another bolt was immunostained with anti-Ydj1 and anti-Pgk1. Pgk1 was used as an internal loading control. Control cultures were transformed with the pHR81GAL vector. (B) Overexpressed Cyc8C caused a dramatic increase in Hsp104 levels. Lysates of Hsp104-GFP cells following overnight overexpression of Cyc8C or the empty vector pHR81 were analyzed by immunoblotting with the indicated antibodies. Pgk1 was used as an internal loading control. (TIF) Table S1 Quantification of Hsp104 levels upon Pin4C overexpression. The Hsp104 level was quantified using ImageQuant software and normalizing against the internal Pgk1 control. The normalized Hsp104 level in cells overexpressing Pin4C was compared with that in cells with the empty vector. Data was presented as mean 6 SD, n = 5. (DOC)