The concept of a prion as an infectious self-propagating protein isoform was initially proposed to explain certain mammalian diseases. It is now clear that yeast also has heritable elements transmitted via protein. Indeed, the “protein only” model of prion transmission was first proven using a yeast prion. Typically, known prions are ordered cross-β aggregates (amyloids). Recently, there has been an explosion in the number of recognized prions in yeast. Yeast continues to lead the way in understanding cellular control of prion propagation, prion structure, mechanisms of de novo prion formation, specificity of prion transmission, and the biological roles of prions. This review summarizes what has been learned from yeast prions.
PRIONS are self-propagating and transmissible protein isoforms. The initial awareness of prions came from mammalian diseases, such as sheep scrapie, human Creutzfeldt-Jacob, and bovine spongiform encephalopathy (“mad cow” disease). These diseases are transmitted by an altered infectious conformational isoform (PrPSc) of a normal cellular protein (PrPc). PrPSc forms cross-β aggregates (amyloids) and is infectious because it captures PrPc molecules and converts them into PrPSc (Prusiner 1982; Colby and Prusiner 2011). More than 30 other human diseases (including Alzheimer’s, Parkinson’s, and Huntington’s diseases) are associated with amyloid formation by various proteins, and for some of them transmissibility has been demonstrated at the cellular level (Aguzzi and Rajendran 2009).
While there is no protein with homology to PrPc in yeast, several yeast proteins have now been shown to exist either in a normal soluble or in a transmissible amyloid form. These different states of the same protein cause distinct phenotypes. Furthermore, each prion protein can generally form different types of infectious aggregates with distinct conformations and distinct associated phenotypes, called prion variants. Since soluble molecules conform to the specific prion variant conformation of the aggregates that they join, normally only one prion variant of a given protein propagates stably in a single yeast culture.
Yeast prions provide a model system for studying mechanisms of amyloid formation and propagation that are applicable to mammalian and human diseases. They also manifest themselves as heritable cytoplasmic elements and, in this way, provide a mechanism of inheritance that operates at the level of protein conformation rather than nucleotide sequence. The growing number of examples of prions and other amyloid proteins indicates that they may have an important biological role.
Evidence for Prions
In 1994, on the basis of new and old data, Reed Wickner (Wickner 1994) proposed that the previously known yeast non-Mendelian heritable [URE3] and [PSI+] elements (Cox 1965; Aigle and Lacroute 1975) are, respectively, prion forms of the Ure2 and Sup35 proteins. This revolutionary hypothesis at once explained why (1) the phenotypes of [URE3] and [PSI+], respectively, correspond to “loss-of-function mutations” in URE2 or SUP35; (2) [URE3] and [PSI+] are inherited in a cytoplasmic fashion; (3) propagation of [URE3] and [PSI+] requires the respective presence of the Ure2 or Sup35 prion domains (without these domains there will be no prion seeds); (4) unlike loss of a virus or other cytoplasmic nucleic acid, loss of [URE3] and [PSI+] is reversible (the prion protein still present in the non-prion cell can reform the prion); and (5) transient Ure2 or Sup35 overproduction, respectively, induces the de novo appearance of [URE3] or [PSI+] (excess prion protein enhances the chance that some of it will misfold to form a prion seed, as described below). A great deal of additional evidence has now accumulated to support this hypothesis.
Prions often have loss-of-function phenotypes
Assuming that the function of the normal cellular protein is compromised to some extent when it forms a prion aggregate, the prion phenotype would reflect this loss of function. As an example, Figure 1 illustrates the phenotypic differences between yeast cells with the non-prion vs. prion forms of the translational termination factor Sup35. In the presence of the prion, the translational termination activity of Sup35 is compromised (loss of function) so cells terminate translation less efficiently at nonsense codons (Cox 1965). Likewise, [URE3] (Aigle and Lacroute 1975) and several other yeast prions (see Prion Diversity and Table 1) cause loss-of-function phenotypes. Ure2, a regulator of nitrogen catabolism, prevents uptake of allantoate (a potential nitrogen source) in the presence of other nitrogen sources, and of ureidosuccinate (USA), an intermediate in uracil biosynthesis, which closely resembles allantoate. Thus inactivation of Ure2 by mutation or by formation of the [URE3] prion allows ura2 mutant cells that cannot synthesize USA to grow on –Ura by taking up USA (Lacroute 1971). [URE3] can also be scored by a variety of other assays that do not require the presence of a ura2 mutation (Moriyama et al. 2000; Schlumpberger et al. 2001; Brachmann et al. 2006; Hong et al. 2011).
In addition to the loss-of-function phenotypes, prions can also gain new functions. For example, as described below (see Heterologous prion cross-seeding), the presence of the prion form of the Rnq1 protein, called [PIN+] (or [RNQ+]) can enhance the chance that another prion will form de novo. Also, the presence of prions can be associated with toxicity (see Biological Effects of Prions).
Prions are inherited in a non-Mendelian fashion
Since prion aggregates capture and convert non-prion protein into the prion conformation, prion traits are dominant and have a chance of being inherited by all meiotic progeny. In addition, since prions seeds are located in the cytoplasm, they are efficiently transferred by cytoduction. Cytoduction entails the fusion of donor and recipient cells’ cytoplasms without nuclear fusion and when the “donor” nucleus is lost (Conde and Fink 1976; Zakharov and Yarovoy 1977). Thus non-prion recipient cells become infected with prion when cytoduced with a prion-containing “donor” cell (Cox 1993).
The fact that isolates of mammalian prion diseases in otherwise genetically identical animals showed different stable and reproducible characteristics was a challenge for the “protein only” model of prion phenomena as this result suggested an explanation by viral mutations (Dickinson et al. 1968; Bruce and Dickinson 1987; Manuelidis 2003). The finding that the [PSI+] prion also had different heritable states termed “variants” (Derkatch et al. 1996) paralleled the observations in mammals. However, yeast prion variants could not be explained by mutations as different variants could be induced by overproduction of the same protein in the same host.
Different [PSI+] variants were associated with inherently different ratios of aggregated vs. non-aggregated Sup35 protein and therefore caused different degrees of loss of function (Figure 2). Thus, in the presence of the ade1-14 nonsense marker, different variants of [PSI+] cause distinct levels of translational readthrough resulting in characteristic levels of growth on –Ade medium and accumulation of red pigment associated with lack of Ade1. [PSI+] variants that have a larger vs. smaller proportion of aggregated Sup35 protein and thus cause more vs. less translational readthrough are, respectively, called strong vs. weak [PSI+] (Zhou et al. 1999; Uptain et al. 2001). When cells containing different variants of the same prion are mated, the prion variant that replicates more quickly and thus is more highly aggregated takes over the population (Bradley et al. 2002; Tanaka et al. 2006). Once a variant is established, it typically appears to be stable (Derkatch et al. 1996; Kochneva-Pervukhova et al. 2001). However, strong variants may rarely appear spontaneously in the weak variant background, and this could be facilitated by chemicals that selectively cure weak but not strong [PSI+] (Shorter 2010).
Interestingly, different mutations in the prion domain have distinct effects on the phenotype of particular [PSI+] variants (Derkatch et al. 1999; King 2001; Disalvo et al. 2011; Verges et al. 2011). Variants of other yeast prions, namely [PIN+] (Bradley et al. 2002) and [URE3] (Schlumpberger et al. 2001), have also been described.
Correspondence between prions and amyloid aggregates
Considerable evidence indicates that the prion form of most proteins is an amyloid aggregate. The prion vs. non-prion forms of these proteins are protease K resistant and are found preferentially in the pellet vs. supernatant fractions of cell lysates (Masison and Wickner 1995; Patino et al. 1996; Paushkin et al. 1996). Fusions of these prion proteins to fluorescent tags are diffuse in non-prion cells, but form punctate fluorescent dots in cells with the corresponding prion (Patino et al. 1996; Edskes et al. 1999). When lysates treated with detergent at room temperature are fractionated by centrifugation or passed through a filter, prion aggregates are precipitated or trapped by the filter because they are detergent resistant (Scherzinger et al. 1999). The detergent treatment dissolves large prion aggregates into polymers that can be separated on an agarose gel. The size range of the polymers found in cells is characteristic for different prion variants (Kryndushkin et al. 2003; Bagriantsev et al. 2006; Liebman et al. 2006). Interestingly, stronger [PSI+] variants have smaller polymers than weaker [PSI+] variants. The reasons for this will be discussed below (see Requirements for Prion Propagation: Shearing and Segregation). Additional in vitro and in vivo evidence that prions form amyloids is described in Models of prion structures.
Transfection of prions
Proof of “protein-only” infection by a prion required that purified prion aggregates added to a cell would cause infection. This was first demonstrated with Sup35 in Saccharomyces cerevisiae (Sparrer et al. 2000) and prion protein HET-s in the fungus Podospora anserina (Maddelein et al. 2002). However, since overexpression of a prion protein even if it is not in the infectious prion conformation will also induce de novo prion appearance at a high frequency (Wickner 1994; Masison and Wickner 1995; Derkatch et al. 1996), it was essential to distinguish infection from de novo induction. Since de novo prion appearance will include a variety of prion variants, the definitive proof required a demonstration that the prion protein infection was variant specific (Liebman 2002).
This was first done simultaneously by two groups. The C. King group (King and Diaz-Avalos 2004) used a tagged Sup35 fragment purified from cells propagating different [PSI+] variants to seed in vitro fiber formation with bacterially expressed Sup35. These fibers, when sheared and introduced into [psi−] cells, reproduced the initial [PSI+] variants (Figure 3). In another version of this experiment, J. Weissman’s group (Tanaka et al. 2004) used a bacterially expressed Sup35 fragment incubated at different temperatures to make fibers with distinct conformations that, when transfected into [psi−] yeast, produced specific variants of [PSI+]. Likewise, Ure2 fibers seeded in vitro with variant-specific [URE3] cell extracts infected [ure-o] cells with the corresponding [URE3] variant (Brachmann et al. 2005). In vitro-made fibers of a number of other yeast prions have also been shown to infect cells with the corresponding prion (see Table 1).
Requirements for Prion Propagation: Shearing and Segregation
Role of Hsp104 in prion propagation
While prion proteins can generate and propagate an amyloid state in vitro in the absence of any other cofactors, in vivo propagation of yeast prions depends on the chaperone machinery. The Hsp104 chaperone, a homohexameric AAA ATPase, is required for the propagation of [PSI+] (Chernoff et al. 1995). Deletion of HSP104 eliminates [PSI+], and dominant negative HSP104 mutations antagonize [PSI+]. Hsp104 is also required for propagation of the other proven amyloid-based yeast prions (see Chernoff 2007; Rikhvanov et al. 2007; Romanova and Chernoff 2009; Reidy and Masison 2011), with the exception of [ISP+] (Rogoza et al. 2010) and, possibly, the [PHI+] prion that is based on an artificially engineered derivative of Sup35 (Crist et al. 2003). The effects of Hsp104 on yeast prions are summarized in Table 2.
Hsp104 and its bacterial ortholog, ClpB, are implicated in disaggregation of stress-damaged proteins (see Glover and Lum 2009). It was proposed that Hsp104 promotes fragmentation of prion fibers into smaller seeds, initiating new rounds of prion propagation (Paushkin et al. 1996; Kushnirov and Ter-Avanesyan 1998). Indeed, a decrease in Hsp104 activity results in the accumulation of larger Sup35 prion aggregates (Wegrzyn et al. 2001; Satpute-Krishnan et al. 2007), composed of longer SDS-resistant polymers (Kryndushkin et al. 2003). Excess Hsp104 promotes, rather than counteracts, propagation of Sup35 prion variants with abnormally large aggregates, generated by altered Sup35 protein (Borchsenius et al. 2001) or selected at high levels of Hsp104 (Borchsenius et al. 2006).
These data are consistent with the view that the crucial role of Hsp104 in prion propagation is prion fragmentation. One possibility is that Hsp104 breaks prion polymers by pulling individual protein molecules from the middle of the polymer through the central pore of the Hsp104 hexamer. This is how Hsp104 and its bacterial ortholog, ClpB (Weibezahn et al. 2004; Glover and Lum 2009), solubilize aggregated stress-damaged proteins. Indeed, a modified version of Hsp104, HAP, or the modified Hsp104-ClpB chimera, 4BAP, which contains a docking site for the inactive bacterial protease ClpP and is able to capture protein molecules pulled from aggregates, can capture Sup35 in [PSI+] cells (Tessarz et al. 2008; Tipton et al. 2008). On the other hand, the location of some Hsp104 mutations that affect prion propagation, but not the solubilization of stress-damaged proteins and thermotolerance, along the lateral channel of the hexamer rather than central pore (see Romanova and Chernoff 2009) suggests that the lateral channel interacts with prions (Kurahashi and Nakamura 2007). Some data indicate that Hsp104 alone can promote fragmentation of prion fibers in vitro (Shorter and Lindquist 2004), while results of other groups disagree (Inoue et al. 2004; Krzewska and Melki 2006). In any case, in vivo effects of Hsp104 are strongly influenced by other chaperones, as discussed below.
Overproduction of Hsp104 causes loss of [PSI+] (Chernoff et al. 1995) but not of the other known prions. It was proposed that excess Hsp104 eliminates [PSI+] by disaggregating prions to monomers (Paushkin et al. 1996; Kushnirov and Ter-Avanesyan 1998). Indirectly, this hypothesis is supported by the observation that Sup35 overproduction, leading to an increase in aggregate size, partly ameliorates the curing effect of excess Hsp104 (Borchsenius et al. 2006). Also, in vitro, a huge excess of Hsp104 leads to the “remodeling” of Sup35-based amyloids, resulting in the loss of their ability to transmit the prion state via transfection, while Ure2-based amyloids retain infectivity (Shorter and Lindquist 2006). Another possibility could be that excess Hsp104 prevents Sup35 monomers from efficiently joining larger polymers. However, the reverse effect is also possible: if Hsp104 overproduction causes an increase rather than a decrease in the size of Sup35 polymers, this might impair prion propagation as well. Indeed, weak [PSI+] variants that are less efficiently fragmented by Hsp104 than by strong [PSI+] under normal conditions (see Role of polymer growth and fragmentation in determining differences between prion variants) are more sensitive to Hsp104 overproduction than to strong [PSI+] (Derkatch et al. 1996). Also, Hsp104 overproduction results in an increase in the size of the remaining Sup35 polymers as soluble Sup35 monomers accumulate (Kryndushkin et al. 2003). However, this latter result could also be explained if the larger molecular weight polymers are retained due to their greater resistance to Hsp104. Finally, since [PSI+] curing by excess Hsp104 requires the N-terminal region of Hsp104 that is not required for prion propagation (Hung and Masison 2006), it appears that Hsp104-mediated shearing is not sufficient for prion curing by excess Hsp104. Overall, the mechanism by which excess Hsp104 antagonizes [PSI+] and the reasons for differential sensitivities of yeast prions to excess Hsp104 remains unclear. One hypothetical model will be discussed below (see Prion segregation at cell division).
Hsp104 is conserved in many organisms other than Saccharomyces, although multicellular animals do not appear to have an orthologous cytoplasmic protein (see Rikhvanov et al. 2007; Romanova and Chernoff 2009). However, mammalian cells do exhibit induced thermotolerance (Li et al. 1995), which is controlled by Hsp104 in other organisms. The ability of Hsp104 to support prion propagation is conserved in some but not all species: Hsp104 from Candida albicans (Zenthon et al. 2006) but not from Schizosaccharomyces pombe (Senechal et al. 2009) supports the propagation of [PSI+] in S. cerevisiae. Intriguingly, the C. albicans Sup35 protein can acquire a prion state in S. cerevisiae cells (Santoso et al. 2000; Resende et al. 2002), while the S. pombe Sup35 protein lacks a prion domain (Ito et al. 1998; Kong et al. 2004).
Role of other Hsps
The yeast Hsp70 and Hsp40 chaperones are also implicated in prion propagation (see Rikhvanov et al. 2007; Romanova and Chernoff 2009; Reidy and Masison 2011). Data on Hsp70 effects are summarized in Table 2. Yeast contains two major cytosolic Hsp70 subfamilies, namely Ssa (working with the Hsp40 co-chaperones Ydj1 and Sis1) and Ssb (working with Hsp40-Zuo1 and Hsp70-related co-chaperone Ssz1). Ssa is encoded by four genes (constitutive SSA2, moderately expressed and stress-inducible SSA1, and strictly stress-inducible SSA3 and SSA4), of which at least one must be present for viability. Ssb, encoded by two genes, SSB1 and SSB2, is nonessential, not heat inducible, ribosome associated, and implicated in folding of nascent polypeptides. Remarkably, Ssa and Ssb exhibit opposite effects on the [PSI+] prion (Chernoff et al. 1999; Newnam et al. 1999; Allen et al. 2005): Ssa overproduction partly protects [PSI+] from curing by excess Hsp104, while Ssb overproduction enhances curing, and deletion of both SSB genes, ssb1Δ ssb2Δ, counteracts curing (Chernoff et al. 1999). Ssa overproduction (Newnam et al. 1999; Allen et al. 2005) or ssb1Δ ssb2Δ deletion (Chernoff et al. 1999) also increases translational readthrough in [PSI+] strains and promotes de novo [PSI+] formation (see De Novo Prion Formation). Experiments with chimeric proteins indicate that the peptide-binding domain of Hsp70 is responsible for the differences in the effects of Ssa and Ssb on [PSI+] (Allen et al. 2005).
Interestingly, overproduction of Ssa may also antagonize [PSI+] propagation, as seen for some [PSI+] variants, formed by altered (Borchsenius et al. 2001) or wild type Sup35 (Borchsenius et al. 2006) that are aided (rather than cured) by excess Hsp104, and for other [PSI+] variants in the presence of overproduced Sup35 (Allen et al. 2005) and/or in the presence of another prion, [PIN+] (Mathur et al. 2009). In the latter case, the [PSI+] curing effect of excess Ssa was linked to an increase in the size of cytologically detectable Sup35 aggregates, leading to decreased transmissibility of these aggregates in mitotic divisions. At a molecular level, excess Ssa increases both the size of Sup35 polymers and the proportion of non-aggregated Sup35 (Allen et al. 2005). Ssa physically interacts with Sup35 (Allen et al. 2005) and was identified as a major non-Sup35 component associated with [PSI+] aggregates in vivo (Bagriantsev et al. 2008). Excess Ssb also antagonizes weak variants of [PSI+] upon prolonged propagation (Kushnirov et al. 2000b; Chacinska et al. 2001) or other [PSI+] variants in the presence of excess Sup35 (Allen et al. 2005).
Mutation in SSA1 was shown to antagonize [PSI+] propagation (Jung et al. 2000), and deletion of SSA2, responsible for the major fraction of Ssa in exponentially growing cells, destabilizes a weak [PSI+] variant (Newnam et al. 2011). Overall, effects of the ssa mutations on [PSI+] resemble effects of Hsp104 overproduction (see Reidy and Masison 2011). It appears that consequences of chaperone action on Sup35 aggregates depend on the balance between Hsp104 and Ssa, rather on the amount of each of these proteins per se. Indeed, [PSI+] is destabilized following short-term heat shock when Hsp104 levels increase more quickly than levels of other Hsp’s including Ssa, while longer incubation at high temperature, resulting in partial restoration of the Hsp104/Ssa balance, leads to restoration of [PSI+] stability (Newnam et al. 2011). Deletions of individual SSA genes increase [PSI+] destabilization by short-term heat shock and impair [PSI+] recovery after longer heat shock, confirming the role of Hsp104/Ssa balance in prion maintenance during and after stress.
Each member of the Ssa family acts on [PSI+] in the same direction when it is overexpressed (Allen et al. 2005); however, differential effects are detected when each Ssa protein is expressed individually in cells lacking Ssa proteins (Sharma and Masison 2008). Ssa proteins also differ from each other in their effects on other prions. For example, [URE3] is cured by Ssa1 overproduction but not by Ssa2 overproduction (Schwimmer and Masison 2002), with a single amino acid change at position 83 being responsible for these differences (Sharma and Masison 2011). Mutation in SSA2 also antagonizes [URE3] (Roberts et al. 2004).
Regarding the effects of Hsp40 proteins, much of our knowledge comes from studying prions other than [PSI+] (see Table 2). The evidence implicating Ydj1 (Moriyama et al. 2000; Sharma et al. 2009; Hines et al. 2011) and Sis1 (see below) in prion propagation is unambiguous. As Sis1 is essential (see Rikhvanov et al. 2007), its effect on prions were studied with mutants, internal deletions, or transient depletions, rather than complete disruptions. Intact Sis1 is required for the maintenance of [PIN+] (Sondheimer et al. 2001) and [URE3] (Higurashi et al. 2008). Sis1 also aids in [PSI+] propagation (Higurashi et al. 2008) and promotes [PSI+] curing by excess Hsp104 (Kirkland et al. 2011). Sup35 capture by the above-mentioned “trapping” derivative of Hsp104, 4BAP, depends upon Sis1 (Tipton et al. 2008). This suggests that Sis1 is responsible for recruiting Hsp104 and possibly Ssa to prion polymers. However, the substrate-binding region of Sis1 is dispensable for [PSI+] propagation, contradicting this model (Kirkland et al. 2011). Whatever the specific mechanisms of interactions, it is clear that members of both Hsp70 and Hsp40 chaperone families, apparently working together with Hsp104, play crucial roles in prion propagation.
Less is known about whether chaperones other than Hsp104, Hsp70, and Hsp40 influence prions. Elimination or overproduction of the Hsp70 nucleotide exchange factors (NEFs) Fes1 or Sse1 (Jones et al. 2004; Kryndushkin and Wickner 2007; Sadlish et al. 2008; Hines et al. 2011) affects [PSI+], [URE3], and [SWI+] prions. Mutation analysis suggests that NEFs act on prions via regulating Ssa. Nucleotide exchange is needed for Hsp70 to release substrates. Thus it seems that Ssa acts on prions by binding and releasing them, just as it binds and releases other misfolded proteins. Chemical inhibition of Hsp82, a yeast counterpart of the Hsp90 chaperone, or deletions of the genes encoding the Hsp70/Hsp90 co-chaperones, Sti1 or Cpr7, counteract [PSI+] curing by excess Hsp104 (Moosavi et al. 2010; Reidy and Masison 2011); however, these effects could be mediated by Ssa, as Hsp82 deficiency increases Ssa levels.
Importantly, Hsp104, Hsp70-Ssa, and Hsp40s (Sis1 and Ydj1) represent the major complex involved in disaggregation and refolding of stress-damaged proteins (Glover and Lindquist 1998; Glover and Lum 2009). Hsp70 and Hsp40 components of this complex are conserved in other organisms, including humans (see Rikhvanov et al. 2007), suggesting that these data likely have implications for mammalian amyloids as well. It is remarkable that the same chaperone machinery is employed in protection against environmental stresses and in modulating amyloid propagation. Apparently, effects of molecular chaperones on prions are based on the same enzymatic activities that are involved in their interactions with other misfolded and/or aggregated proteins. However, the highly ordered nature of prion aggregates increases their resistance to the Hsps’ action and therefore alters consequences of the aggregate/Hsp interaction. Instead of eliminating an aggregate, Hsps (at least at certain levels or, more likely, in certain ratios) promote fragmentation, which multiplies aggregated seeds and thus facilitates prion propagation. Such a unique response to chaperone action makes aggregates capable of behaving in a prion fashion in vivo. Therefore, it is this specific mode of interaction with the chaperone machinery that makes a yeast protein aggregate a prion.
Role of polymer growth and fragmentation in determining differences between prion variants
During prion propagation, mature protein molecules (rather than only newly synthesized ones) can be remodeled to join pre-existing prion polymers (Satpute-Krishnan and Serio 2005). Efficiency of polymer fragmentation by chaperones relative to polymer growth explains phenotypic differences between prion variants (Tanaka et al. 2006). Polymers of strong [PSI+] variants are readily fragmented and therefore produce a larger number of prion units per cell. As termini of prion polymers are active in attracting new protein molecules to the polymers, a larger number of polymers results in the more efficient immobilization of newly synthesized Sup35 protein into polymers. In contrast, polymers of weak [PSI+] variants are less efficiently fragmented, resulting in fewer polymer ends and less efficient capture of new Sup35 molecules (see also Figure 2D). This explains why weaker [PSI+] variants are characterized by a larger average polymer size (Kryndushkin et al. 2003) and a higher proportion of non-aggregated Sup35 protein (Zhou et al. 1999; Uptain et al. 2001), leading to a less severe defect in termination (Derkatch et al. 1996) when compared to stronger [PSI+] variants. Sensitivity to chaperones depends on the physical properties of polymers controlled by the conformation of the prion aggregate. Apparently, a large amyloid core (Toyama et al. 2007) makes weak prion variants more physically stable and less accessible to Hsps than strong prion variants with a smaller amyloid core. Thus, biologically weak prion variants (as judged from phenotype and transmissibility) are based on polymers that are physically stronger. Amyloids that are absolutely stable and rigid in vivo would not produce new “seeds” and therefore would not be expected to behave as prions.
Prion segregation at cell division
The mitotic stability of prions requires that prions segregate to daughter cells. Guanidine hydrochloride (GuHCl), a compound that blocks prion propagation, has been employed to analyze prion segregation in mitosis. GuHCl was initially described as an antagonist of [PSI+] (Tuite et al. 1981; Cox et al. 1988) and was later shown to antagonize all other known amyloid-based yeast prions (reviewed in Cox et al. 2007; Halfmann and Lindquist 2010). The effect on [PSI+] is best understood. [PSI+] curing by millimolar concentrations of GuHCl occurs only in proliferating cells (Eaglestone et al. 1999; Byrne et al. 2007). GuHCl neither prevents Sup35 aggregation nor destroys aggregates (Ferreira et al. 2001; Ness et al. 2002). Rather, it blocks the fragmentation of existing prion units, thereby preventing the generation of new prion units. This leads to the dilutions of prion units as the cells divide, and eventually daughter cells do not inherit any prion units. GuHCl antagonizes Hsp104-induced thermotolerance in vivo (Ferreira et al. 2001; Jung and Masison 2001) and inhibits the ATPase activity of Hsp104 in vitro (Grimminger et al. 2004), suggesting that its effect on prions is also primarily due to inhibition of Hsp104. This was confirmed by the identification of a mutation in Hsp104 that makes [PSI+] much less sensitive to the curing effect of GuHCl (Jung et al. 2002). However, differences in kinetics of [PSI+] loss in the presence of GuHCl and after direct Hsp104 inactivation by genetic manipulations (Wegrzyn et al. 2001; Chernoff 2004b) suggest that the picture could be more complex. Indeed, the [ISP+] prion, which does not require Hsp104 for its propagation, is curable by GuHCl (Rogoza et al. 2010). Thus it appears that GuHCl also acts on other targets influencing prion propagation, in addition to Hsp104. Whatever the molecular specifics of GuHCl action, its ability to block the generation of new proliferating prion units (termed “propagons”) can be used to count the number of propagons in a yeast cell.
The number of propagons in a cell can be derived from the number of cell divisions needed for prion loss in the presence of GuHCl or by determining the number of cells that retain a [PSI+] seed in a colony derived from a single [PSI+] cell grown in the presence of GuHCl (Cox et al. 2003, 2007). One caveat with these methods is that seeds are preferentially retained by mother cells (Byrne et al. 2009). While cell-to-cell variation in propagon numbers (from several to more than a thousand per cell) was uncovered in yeast cultures, strong prion variants are characterized by a larger average number of propagons per cell, compared to weak prion variants. This agrees with the fragmentation model (see above, Role of polymer growth and fragmentation in determining differences between prion variants) and accounts for differences in mitotic stability.
Yeast cultures bearing a weak [PSI+] variant exhibit asymmetric accumulation of larger prion polymers in aged cells (Derdowski et al. 2010). It was proposed that larger polymers are less likely to be transmitted to a daughter cell (bud) during mitosis. Loss of weak [PSI+] in the first cell division after heat shock (Newnam et al. 2011) also preferentially occurs in daughters. The size of prion polymers is increased during heat shock, and this may contribute to the asymmetry. However, it seems unlikely that increased polymer size per se represents a mechanical threshold for transmission, as even much larger intracellular structures are transmitted from the mother cell to the bud. It is worth noting that non-prion protein aggregates produced during heat shock, e.g., agglomerates of oxidatively damaged proteins, are also preferentially accumulated in the mother cell (Aguilaniu et al. 2003; Liu et al. 2010). Hsp104 binds these agglomerates and plays a crucial role in their mother-specific accumulation (Erjavec et al. 2007; Tessarz et al. 2009). Decreased diffusion of the larger aggregates through the budneck, coupled with more efficient solubilization of aggregates in the bud, was suggested as an explanation for the asymmetry (Zhou et al. 2011). However, other evidence indicates that at least some aggregates either are trapped in a scaffold of actin cables in the mother cell or are subject to active retrograde transport back to the mother from the growing bud, involving the polarisome and the actin cytoskeletal network (Liu et al. 2010, 2011). Cytoskeletal structures are linked to prion segregation as well. For example, weak variants of [PSI+] are destabilized after prolonged disruption of actin cytoskeleton by latrunculin A (Bailleul-Winslett et al. 2000), and deletion of the gene coding for actin assembly protein Lsb2 increases [PSI+] destabilization by heat shock (Chernova et al. 2011). Overexpression of Btn2 or Cur1, yeast homologs of mammalian microtubule-associated Hook proteins involved in organelle transport (Walenta et al. 2001), cures [URE3] prion, possibly by impairing its segregation (Kryndushkin et al. 2008).
One possibility is that Hsp104, working in a stoichiometric combination with Ssa and its Hsp40 co-chaperones, fragments prion polymers in vivo, while Hsp104 in imbalance with Ssa directs the association of prion polymers with the cytoskeletal networks, resulting in the mother cell-specific retention and/or retrograde transport. Thus interplay between polymer fragmentation, diffusion into the daughter cell, retention by the mother cell, and/or retrograde transport back to the mother cell regulates prion segregation. In this model, Hsp104 promotes the retention and/or retrograde transport of aggregates when it cannot break them efficiently. Therefore, larger polymers that are less sensitive to Hsp104-mediated breakage would be more likely to be accumulated in the mother cell. Such a process would be adaptive as it protects daughter cells from aggregates at the expense of the aged mothers. It is possible that the same mechanism contributes to [PSI+] curing by plasmid-mediated overproduction of Hsp104 (see above) and that the N-terminal domain of Hsp104, required for curing, is involved in the interactions promoting prion retention and/or retrograde transport. Further experiments are needed to prove or disprove this model. The role of Hsp104 in [PSI+] propagation is summarized on Figure 4.
Structural Organization of Prions
Yeast prion proteins contain regions, termed “prion domains” (PrDs), that are required for formation and propagation of the prion state and can maintain the prion state even without the rest of the protein (Ross et al. 2005b; Inge-Vechtomov et al. 2007). When the major cellular function of the prion protein is known, PrD is typically dispensable for this function. However, PrDs may have functions other than prion formation; e.g., Sup35 PrD (Sup35N) is implicated in interactions with poly(A)-binding protein, influencing mRNA stability (Hosoda et al. 2003). Interestingly, Sup35 also contains a middle region (Sup35M) linking PrD to the C-proximal release factor domain. Sup35M is enriched in charged residues and is suspected of helping to maintain a balance between aggregated and non-aggregated states, possibly via interaction with Hsp’s (Liu et al. 2002). Indeed, Sup35M interacts with Hsp104 in vitro and is involved in [PSI+] curing by excess Hsp104 in vivo (Helsen and Glover 2012). Yeast PrDs may confer a prion state to a different protein when fused to it artificially. Features of some yeast PrDs are shown in Figure 5.
Generally, the yeast prion PrDs known to date are intrinsically disordered in solution and QN-rich. Typically, they are more N- than Q-rich (e.g., Alberti et al. 2009), and “minimal” PrDs may contain no Qs (Crow et al. 2011). Substitution of Qs for Ns increases, while substitution of Ns for Qs decreases, prion propagation by a given protein (Halfmann et al. 2011). “Scrambled” PrDs of Ure2 or Sup35, maintaining amino acid composition but not exact sequence, are typically capable both of generating amyloid in vitro [albeit with altered rates (Liu et al. 2007)] and prion in vivo and of propagating the prion state, indicating that amino acid composition plays the primary role in prion properties (Ross et al. 2004, 2005a). Mutational analysis of a short amino acid stretch within a certain “scrambled” Sup35 PrD suggested that prion propagation propensity could be increased by exclusion of “order-promoting” residues (even if they have amyloidogenic potential) and enrichment with “disorder-promoting” residues (Toombs et al. 2010). The universality of these rules is still to be determined. The N-proximal PrD region of S. cerevisiae Sup35 includes an N-terminal QN-rich stretch, located within the first 40 amino acids, and a region of 5.5 imperfect oligopeptide repeats (ORs), which somewhat resembles repeats of mammalian PrP and are located between positions 41 and 97. The PrD fragment required for aggregation is shorter than the fragment needed for efficient propagation of the prion state (Borchsenius et al. 2001) and is primarily confined to the QN-rich stretch (Osherovich et al. 2004). It was proposed that the Sup35 (as well as New1) PrD can be divided into “aggregation” (QN stretch) and “propagation” (ORs) elements (Chernoff 2004a; Osherovich et al. 2004) and that the propagation element is involved in interaction with Hsp104 (see Requirements for Prion Propagation: Shearing and Segregation). OR expansion increases de novo [PSI+] generation (Liu and Lindquist 1999) although Ure2 or “scrambled” Sup35 PrDs lack ORs, indicating that ORs are not necessary for interaction with the chaperones responsible for prion propagation (Ross et al. 2005b; Toombs et al. 2011). Perhaps ORs are frequently associated with prions because the duplication events that generate them also extend the size of the regions with the amino acid compositions conducive to prion formation.
In the case of the Rnq1 protein, four QN-rich stretches were found within the PrD (Kadnar et al. 2010). While none of these stretches was essential for prion propagation, two of four stretches were each shown to support prion maintenance if retained alone. Multiple stretches exhibit a cooperative effect on prion maintenance: one stretch, not capable of maintaining the prion state on its own, was needed to propagate some but not the other [PIN+] variants, and this was confirmed by mutagenesis experiments (Bardill and True 2009; Stein and True 2011). The mosaic organization of Rnq1 PrD confirms that different sequence elements contribute to prion properties.
Models of prion structures
To fully understand how prions form and propagate, we must know the structure of both the prion and the non-prion states of the protein. In addition, an appreciation of how prion variants arise from different heritable structures requires a comparison of these different structures. Here we concentrate on efforts to determine the structure of protein aggregates in their prion state.
Unfortunately, the traditional approaches of X-ray crystallography and solution NMR are not suitable to solve the structure of prions because their filamentous nature prevents them from forming crystals, and they are too large to be soluble. Instead, solid-state NMR (e.g., Shewmaker et al. 2006), H/D exchange (e.g., Toyama et al. 2007), electron paramagnetic resonance (Tanaka et al. 2004, 2005), and fluorophore labeling (Krishnan and Lindquist 2005) have been used to investigate the structure of yeast prions.
Considerable controversy exists on this topic. The following reviews favor different prion structural models (Tessier and Lindquist 2009; Bockmann and Meier 2010; Wickner et al. 2010; Wickner et al. 2011).
Divide and conquer—determining the structure of prion and non-prion domains separately:
The task of determining the structure of prions is made easier if one assumes that the prion and non-prion domains do not have a major effect on each other’s structure. Indeed, to date, most studies have separately investigated the structures of prion and non-prion domains (Bousset et al. 2001a,b; Umland et al. 2001; Tanaka et al. 2004, 2005; Krishnan and Lindquist 2005; Shewmaker et al. 2006, 2009a; Toyama et al. 2007; Wickner et al. 2008a; Chen et al. 2009; Engel et al. 2011). However, this approach has been questioned because Cys scanning and disulfide bond data suggest that the prion and non-prion domains of Ure2 interact when Ure2 fibers are made under native conditions (Fay et al. 2005). Also, the fact that mutations in non-prion domains can affect prion generation and propagation suggests that there might be some interaction between the domains (Masison and Wickner 1995; Maddelein and Wickner 1999; Shibata et al. 2009; Chen et al. 2011; Kabani et al. 2011; Kurahashi et al. 2011).
Despite these observations, interactions between prion and non-prion domain regions appear to have minor, if any, effects on their structures. Indeed, non-prion domains have been shown to retain their structure and activity even within the prion aggregate (Baxa et al. 2002, 2004, 2011; Bai et al. 2004; Krzewska et al. 2007; Zhang et al. 2008; Zhang and Perrett 2009). More importantly, amyloid fibers of PrD fragments are infectious when transfected into non-prion-containing cells (King and Diaz-Avalos 2004; Tanaka et al. 2004; Brachmann et al. 2005; Diaz-Avalos et al. 2005; Patel and Liebman 2007; Du et al. 2008, 2010; Alberti et al. 2009). Although it has been hypothesized that the fibers induce prion formation de novo by titrating away chaperones rather than by seeding (Bousset et al. 2010), retention of specific prion variants after transfection indicates that the in vitro-generated PrD material is in the infectious prion conformation. Studies demonstrating that the introduction of even a single Sup35NM fiber into a cell can cause the appearance of [PSI+] (see figure S1 in Tanaka et al. 2006) are consistent with the idea that fibers, rather than some contaminating PrD structures, are the infectious material. This validates the relevance of the models derived from studies of PrD amyloid fibers.
Yeast prions are amyloid:
In vitro data:
A purified prion protein or fragments containing its PrD often form mixtures of amyloid fibers with various morphologies (Glover et al. 1997; Diaz-Avalos et al. 2005). Some fibers are twisted, others straight, and the twists have different radii and stiffnesses. These fibers bind dyes indicative of amyloid and have the 4.7-Å X-ray diffraction reflection characteristic of a cross-β structure and diagnostic of amyloid (Glover et al. 1997; King et al. 1997; Schlumpberger et al. 2000; Kishimoto et al. 2004; Baxa et al. 2005; Castro et al. 2011). There is also some evidence that Sup35 prion domain fibers seeded by different [PSI +] variants have slightly different mass per unit length, all hovering around one molecule per 4.6 Å (Diaz-Avalos et al. 2005). Indeed, a larger core region (residues 2–73) was solvent-protected in fibers made at 37° (corresponding to a weak [PSI+] variant) vs. a smaller core for fibers made at 4° (corresponding to a strong [PSI+] variant) (Toyama et al. 2007). Recently, fibers in the weak vs. strong [PSI+] conformation have been shown to have an approximately two-fold difference in average stiffness (Castro et al. 2011). Thus it appears that fiber properties differ for different prion variants.
Using cryo-electron microscopy (cryo-EM) and scanning transmission EM, both Ure2 and Sup35-1-61-GFP fibers were found to have a central small core with globular appendages. Fibers made of only PrD lack the globular appendages. When fibers made of full-length protein were digested with protease, the core region that remained undigested corresponded to the PrD (Baxa et al. 2003; Baxa et al. 2005; Diaz-Avalos et al. 2005; Kryndushkin et al. 2011a). Recently, Sup35 fibers have been found to have an ∼8-nm core with a 65-nm cloud of the globular C-region domain extending far enough from the core to allow the C domain to interact with the ribosome (Figure 6) (Baxa et al. 2011).
Despite the above evidence, controversy over whether the prion form of Ure2 is amyloid remains (Bousset et al. 2002b, 2004; Ripaud et al. 2004; Fay et al. 2005; Redeker et al. 2007). A minority view is that the Ure2 globular domain is part of the fiber core because this domain is more ordered in in vitro Ure2 fibers made of full-length Ure2 vs. the PrD alone (Loquet et al. 2009).
In vivo data:
While it is more difficult to prove the existence of amyloid inside cells, fiber-like structures resembling amyloids have been detected in vivo. Ure2 fibers were identified in [URE3] cells overexpressing Ure2 with EM (Speransky et al. 2001). [PSI+] and [PIN+], but not [psi−] or [pin−], cells were shown to be stained by the dye thioflavin-S (Kimura et al. 2003) that binds amyloid. EM analysis of Sup35 polymers isolated from [PSI+] lysates showed them to be composed of ∼20-nm-wide barrels and other larger structures (Bagriantsev et al. 2008). The fluorescent rings and dots formed in the process of prion induction by overexpressed Sup35-GFP (see Prion induction by overproduction) were shown to be made of fibrils (Tyedmers et al. 2010). Also, fibrils that look like those formed in vitro have been seen in [PSI+] cells by EM in large dot-and-line aggregates as well as in diffuse structures in the cytoplasm (Kawai-Noma et al. 2010).
Parallel in-register β-sheets:
Since amyloids were known to be composed of β-sheets, the finding that scrambling the amino acid sequence of Sup35 and Ure2 PrDs did not destroy their ability to form a prion (Ross et al. 2004, 2005a; Toombs et al. 2010) led Wicker and associates (Ross et al. 2005b) to propose that the prion structures were parallel in-register β-sheets. According to this model, the β-sheets in the PrD of each molecule are aligned with identical residues stacked on top of each other. This forms the amyloid core with the globular non-prion domains hanging off the core. The model nicely explains the data because all the PrD molecules of the same scrambled version would contain the identical scrambled sequence, so all amino acids that favor β-structures would still be available to align and form parallel in-register β-sheets (Figure 7A).
Indeed, several mass-per-unit-length measurements of fibers containing the Sup35 and Ure2 PrDs indicate about one molecule per 4.7 Å as predicted by the stacked architecture of the β-sheets in the parallel in-register model (Baxa et al. 2003, 2011; Diaz-Avalos et al. 2005; Chen et al. 2009). The final evidence in support of this model for yeast prions comes from solid-state NMR data for in vitro-generated infectious fibers of Sup35NM, Rnq1 PrD, and Ure2 PrD and fibers made of Ure2 PrDs with shuffled sequences. The method was to specifically label one or a few amino acids with 13C and to then measure the distance to the nearest labeled residue on a different molecule. For a parallel in-register β-sheet, this measurement will be 4.7 Å (the distance between the β-strands as mentioned above). For any other type of β-sheet, the distances will be larger. One difficulty with this approach is that the number of residues that can be specifically labeled is limited because PrDs are so rich in glutamines and asparagines. Nonetheless, most of the residues examined were within the 4.7-Å distance of the identical residue on a different molecule, strongly supporting the parallel in-register model (Chan et al. 2005; Fayard et al. 2006; Baxa et al. 2007; Shewmaker et al. 2008, 2009b; Wickner et al. 2008a,b; Chen et al. 2009).
A given prion domain is hypothesized to form several parallel in-register β-sheets interspersed with non-β-sheet loops. These non-β-sheet loops can account for the residues that are not within the 4.7-Å distance (Figure 7B). Also, the different β-sheets are proposed to interact with each other to form a “steric zipper” (Figure 7B, bottom) in which the side chains of the residues in the opposing β-sheets interdigitate, forming tight van der Waals bonds named “steric zippers.” Such steric zippers have been seen for crystals of short peptides made of amyloid sequences, including peptides from Sup35 PrD (Nelson et al. 2005; Sawaya et al. 2007; Van Der Wel et al. 2007).
The lengths of the β-sheets and loops are proposed to differ in, and be the basis for, differences between prion variants. Indeed, Sup35NM prion variants formed in vitro differ in the length of the region protected from H/D exchange, which likely corresponds to the β-rich amyloid core (Toyama et al. 2007). Larger regions were protected in fibers formed at 37° (weaker prion variant) compared to fibers formed at 4° (stronger prion variant). This agrees with the higher physical stability of weaker vs. stronger prion variants (see above, Requirements for Prion Propagation: Shearing and Segregation). Once a fiber forms with a set of β-sheets, steric zippers, and loops that represent a particular prion variant, new monomers that join the fiber are expected to be templated to form the same β-sheets, steric zippers, and loops (Figure 7C). The inclusion of different PrD segments into different components of the structure may explain the different effects of specific PrD structural elements on Rnq1 prion propagation (Bardill and True 2009; Kadnar et al. 2010) (see Prion domains) and on the specificity of [PSI+] prion transmission (Chen et al. 2010) (see below, Sequence Specificity of Prion Transmission and Transmission Barriers).
One concern with the solid-state NMR data are that the widths of the lines (each of which represent an atom in a particular environment in the aggregates) of the Sup35, Rnq1, and Ure2 PrD spectra were much broader than expected. This suggests either that the samples are composed of a mixture of fibers with similar but different conformations (possibly a mixture of different prion variants) or that there is some disorder in the fibers, e.g., “breathing” on their ends giving rise to non-β-sheet loops of different sizes (Bockmann and Meier 2010).
More support for the parallel in-register β-sheet model has recently appeared from a study of Ure2 prion domain fibers using site-directed spin labeling and electron paramagnetic resonance (Ngo et al. 2011). This study also provides evidence that a portion of the β-sheet region is more solvent-protected than the rest, suggesting that the β-sheets are organized in inner and outer cores that may differ in different prion strains.
Other in vitro evidence supports a β-helix model for Sup35 PrD (Kishimoto et al. 2004; Krishnan and Lindquist 2005). According to this model, each rung of the β-helix surrounds an empty central cavity (Figure 8).
Krishnan and Lindquist (2005) labeled Cys residues, which they introduced throughout the Sup35NM sequence and which did not alter prion function, with fluorescent dyes responsive to solvent exposure. The solvent-protected core identified by this approach encompassed some (residues 36–86) or most (residues 21–121) of the N domain, depending upon whether the fibers were primarily of the strong [PSI+] variant (made at 4°) or primarily of the weak [PSI+] variant (made at 25°), respectively. The core domains defined by this method are shorter than the region predicted to be part of the Sup35NM parallel in-register β-sheets. Even shorter core regions were deduced from H/D exchange data (Toyama et al. 2007). However, the parallel in-register structure could in principle be reconciled with these results if the edges of the β-sheet domains dynamically expand and contract. This “breathing” might prevent the β-sheet domains from being solvent-protected, but still allow the detection of weak intermolecular self-interactions with solid-state NMR.
According to the β-helix model, only rungs at the top (head) and bottom (tail) of the solvent-protected region would have intermolecular contacts. Indeed, peptide array experiments have identified sites within head and tail regions of the Sup35 PrD as primary sites of intermolecular interactions (Tessier and Lindquist 2007), although it is not clear if interactions uncovered by this approach are identical to those involved in amyloid formation. By labeling individual Cys residues with fluorophores that respond to the presence of nearby dye, Krishnan and Lindquist (2005) detected intermolecular interactions only between residues located in the head and tail regions, and not between residues in the central PrD region. To address the concern that the large fluorophores might alter the prion structure, the authors demonstrated that disulfide bonds between Cys residues in the head region or in the tail region enhanced or did not alter the rate of amyloid formation, while disulfide bonds in the central region were inhibitory. These data could also be consistent with the parallel in-register β-sheet model if the Cys residues in the central region fell within a non-β-sheet loop. Likewise, the finding that interactions in the head and tail regions are crucial for initiating amyloid aggregation is consistent with both models. However, there is no easy explanation for the faithful reproduction of prion variants by the β-helix model, as in this model newly joining PrD initially interacts with the pre-existing structure only at one end.
An important clue to distinguish between the β-helix and parallel in-register models is the 8- to 10-Å reflection in the X-ray diffraction pattern, which is predicted only by the parallel in-register β-sheet model. While this reflection is generally agreed to be present in dried fibers, it has been reported to be missing in hydrated fibers, suggesting that the dried fibers and hydrated cellular prion could be in different confirmations (Bousset et al. 2002a, 2003, 2004; Fernandez-Bellot et al. 2002; Kishimoto et al. 2004). However, two groups have found this reflection to be associated even with hydrated prion fibers (Shewmaker et al. 2009a; Wang et al. 2011).
So far, all structural data for yeast prions has been obtained with in vitro-generated fibers, and no approach has produced a structure at atomic resolution. Also, only one or two variants have been studied in each set of experiments. Among non-yeast amyloids, there are examples of both parallel in-register β-sheets (e.g., human Aβ, amylin, and α-synuclein) and possible β-helices (e.g., bacterial curli) (Shewmaker et al. 2009b). It is very possible that different yeast prions, or even different variants of the same prion, may have very different structures. Indeed, a mutation in Aβ can lead to the formation of a predominantly antiparallel, rather than a typical parallel, β-sheet (Tycko et al. 2009).
Structure of a well-defined fungal prion, [HET-s]:
The [HET-s] prion of the fungus P. anserina (Saupe 2011) is the only prion whose structure is known at the level of atomic resolution (Van Melckebeke et al. 2010). Although the Het-s PrD is not QN-rich, there are a lot of other similarities with the yeast prions. Het-s fibers have an amyloid core with globular appendages (Ritter et al. 2005; Sen et al. 2007). The core is made of the PrD and is protease resistant and infectious (Maddelein et al. 2002), supporting the globular decoration model. The Het-s PrD structure is the same whether fibers are made of only PrD or of the complete protein (Wasmer et al. 2009). Unlike the solid-state NMR data for the yeast prions, the data for the Het-s PrD contain very narrow bands, indicative of a single structure with little disorder (Bockmann and Meier 2010; Van Melckebeke et al. 2010). This may be because there are no variants of the [Het-s] prion; indeed, no variants have been reported.
The [Het-s] prion domain structure combines elements of both the β-helix and the parallel in-register β-sheet models. It has modified parallel in-register β-sheets in the shape of a left-handed β-solenoid that surround an empty central cavity. There are two windings per molecule leading to a mass per unit length of one molecule per 9.4 Å (Sen et al. 2007; Mizuno et al. 2011) rather than the 4.7 Å seen for the yeast prions. There are eight β-strands per molecule (Figure 9). Strands 1a and 3a, 1b and 3b, 2a and 4b, 2b and 4b are pseudodirect repeats in amino acid sequence that align with their pseudorepeat partner in parallel and in register. Additional molecules align so that all the pseudorepeat β-strands form parallel in-register sheets. Three of these sheets define a hydrophobic triangular core while the fourth points away from the core. The two β-sheet layers per molecule are connected by a flexible linker. As in globular proteins, hydrophobic residues are found pointing into the core while polar residues are on the surface.
De novo Prion Formation
Prion induction by overproduction
Transient overexpression of a variety of prion proteins has been shown to dramatically increase (in some cases, as much as 3000-fold or more) the chance that the overexpressed protein will form a prion seed de novo (Chernoff et al. 1993; Wickner 1994; Derkatch et al. 1996; Alberti et al. 2009; Patel et al. 2009). Indeed, transient overexpression of just a PrD (or a portion of PrD) can cause this effect and is often more effective than overproduction of the entire protein. One reason overproduction could induce prion formation is that the increase in protein level could make it more likely for misfolding events to occur, e.g., because of an insufficient supply of chaperones. At higher local concentration it would also be easier for monomers to find each other and aggregate. PrDs may also be more likely to misfold when they are not in the context of the complete protein. Also, the increased protein levels may cause misfolded protein to escape degradation by proteolytic pathways.
Still, overproduction per se is not always sufficient for prion formation. For some prions, the frequency of prion induction by transient overproduction changes dramatically, depending upon the presence of other (heterologous) prions or prion-like aggregates. The best-studied and most dramatic example of this is the induction of [PSI+], which is greatly enhanced by the presence of the [PIN+] prion, other QN-rich prions, or QN-rich proteins in an aggregated state (Derkatch et al. 1997, 2001; Derkatch and Liebman 2007). [PIN+] also enhances the de novo appearance of [URE3], although effects are much less dramatic (Bradley et al. 2002), and increases the induction of the non-QN rich Podospora prion [Het-s] about twofold in yeast (Taneja et al. 2007). When [PSI+] was induced by Sup35 overproduction at a lower frequency in a [psi− pin−] background, each [PSI+] cell was shown to have also picked up a de novo-formed [PIN+] prion that likely facilitated [PSI+] appearance (Derkatch et al. 2000). However, the [PIN+] prion per se is not required for [PSI+] formation, as [PSI+] could also form de novo even in strains that lack the [PIN+]-forming protein, Rnq1 (Chernova et al. 2011; Z. Yang, J. Hong, I. L. Derkatch, and S. W. Liebman, unpublished results). Since, as explained below, other prions or prion-like aggregates may substitute for [PIN+], it is possible that another aggregate helped [PSI+] to appear in these cases.
Heterologous prion cross-seeding
[PIN+] was first identified as a non-Mendelian factor that enhanced the appearance of [PSI+] and had prion-like properties (Derkatch et al. 1997, 2000) and was then shown to be a prion form of Rnq1 (Derkatch et al. 2001). A separate study identified Rnq1 as a prion-forming protein on the basis of a similarity of amino acid composition to Sup35 PrD (Sondheimer and Lindquist 2000; Sondheimer et al. 2001). Since rnq1Δ strains did not enhance [PSI+] induction, it was clear that the [PIN+] prion phenotype was not due to inactivation of Rnq1. Furthermore, other prions or overexpression of other QN-rich proteins did confer the Pin+ phenotype to yeast cells (Derkatch et al. 2001, 2004; Osherovich and Weissman 2001; Meriin et al. 2002). This led to the hypotheses that the [PIN+] prion might (1) titrate away cellular factors that inhibit [PSI+] prion formation and/or (2) provide an initial nucleus to cross-seed the de novo prion aggregation of the heterologous Sup35 QN-rich protein (Derkatch et al. 2001; Osherovich and Weissman 2001) (see Figure 10). However, candidates for [PIN+]-sequestered factors that inhibit prion formation have not been identified despite several genetic screens (Osherovich and Weissman 2002). On the other hand, there is significant evidence in support of the cross-seeding model. Purified Rnq1 PrDs can form fibers in vitro (Glover et al. 1997; King et al. 1997; Patel and Liebman 2007; Vitrenko et al. 2007), and the presence of these fibers enhances the fibrillization of Sup35 PrD and vice versa (Derkatch et al. 2004; Vitrenko et al. 2007). Likewise, yeast Sup35 PrD overexpressed in bacteria formed amyloid fibers, but only if another QN-rich amyloid was already present (Garrity et al. 2010). In addition, cross-seeding can be imitated artificially by fusing Sup35 (or Ure2) PrDs to Rnq1 PrD: such fusions induced [PSI+] (or [URE3]) even when expressed only at a low level, but this was completely dependent upon [PIN+] (Choe et al. 2009). Also mutations in the Rnq1 prion domain that specifically alter the ability of [PIN+] to promote the appearance of [PSI+] have been isolated (Bardill and True 2009).
The de novo induction of [PSI+] by transiently overproduced Sup35 (or its PrD) in the presence of [PIN+] goes through several stages (Figure 11). First, amyloid-like detergent-resistant Sup35 polymers accumulate (Salnikova et al. 2005). When Sup35-based constructs are fused to GFP, the aggregated protein is initially seen as large filamentous (ring-like) assemblies at the cell periphery. The appearance of peripheral rings is increased in nondividing cells, and these rings are transient: they later collapse into smaller internal rings that surround the vacuole. Finally, cells with rings give rise to daughter cells with dot-like aggregates, characteristic of [PSI+] (Zhou et al. 2001; Ganusova et al. 2006; Mathur et al. 2010). If fluorescently tagged Sup35 is overproduced in a cell with established [PSI+], the typically detectable numerous small prion aggregates, distributed throughout the cell, are replaced by one or a few large dots or clumps, but rings do not appear. Remarkably, similar induction aggregate morphologies have been observed for other prions (Derkatch et al. 2001; Alberti et al. 2009), including the non-QN-rich Podospora prion, [Het-s], when induced in yeast (Mathur et al. 2010). Peripheral Sup35-GFP rings overlap cortical actin patches, and both peripheral and internal rings accumulate actin assembly proteins (e.g., Sla1 and Sla2) (Ganusova et al. 2006). This suggests a role for the actin cytoskeleton in ring assembly and/or collapse. Indeed, alterations of the actin cytoskeleton influence both ring formation and [PSI+] induction by excess Sup35 as described below. It was hypothesized that actin cytoskeletal networks assemble misfolded proteins into quality control deposits and, in this way, promote initial prion formation (Ganusova et al. 2006).
Ring structures followed by dots also appear when Sup35NM-GFP is constitutively overproduced, e.g., with a GPD promoter (Tyedmers et al. 2010). EM indicated that, while both the rings and prion dots are composed of fiber bundles, the fibers in rings are very long while those in dots are highly fragmented. Furthermore, lysates of cells with rings can efficiently infect cells with [PSI+] if the fibers are sheared. Rings detected during constitutive overproduction overlapped the preautophagosomal markers, characteristic of the insoluble protein deposit (IPOD). In cells with proteolysis defects, IPOD is formed as a single vacuole-associated structure that collects irreversibly aggregated and oxidatively damaged proteins (Kaganovich et al. 2008). Rnq1 has been found in IPOD, although it is not known if Rnq1 was in the prion state in the cells used for this study. It was proposed that [PSI+] formation is initiated at the IPOD (Tyedmers et al. 2010). However, it is not yet clear if a single IPOD is present in nonstressed cells. It is possible that dispersed preautophagosomal proteins first interact with cytoskeleton-associated aggregates (and/or Rnq1), while formation of IPOD-like structures occurs at subsequent stages. The continuous overexpression of Sup35NM-GFP in the Tyedmars et al. (2010) study makes it possible that the majority of cells examined inherited Sup35 rings from the previous generation, making such persistent rings different from the peripheral rings formed de novo when Sup35NM-GFP is overproduced transiently.
It is necessary to note that the de novo induction of some prions, e.g., [MOT3+], by PrD overproduction does not appear to depend on the presence of pre-existing prions (Alberti et al. 2009). For other prions, the requirement for a pre-existing nucleus for their efficient de novo formation can be overcome. Indeed, overproduction of certain Sup35 PrD-containing fragments with a short extension of hydrophobic residues (Derkatch et al. 2000) or with an addition of the expanded polyglutamine stretch (Goehler et al. 2010) allows efficient de novo induction of [PSI+] in [pin−] cells.
Spontaneous and environmentally induced prion formation
The rate with which prions appear de novo without infection or overexpression varies dramatically for different prions and is very dependent upon the environment. [PIN+] enhances not only [PSI+] induction by overproduced Sup35 (or Sup35 PrD) but also the spontaneous appearance of [PSI+] in the absence of overproduction (Derkatch et al. 1996, 1997, 2000, 2001).
Rates of spontaneous [PSI+] formation have been difficult to measure because Mendelian nonsense suppressors that mimic the [PSI+] phenotype arise at a higher frequency than bona fide [PSI+]. However, a fluctuation test combined with determining the proportion of GuHCl curable (i.e., prion) colonies among all suppressor colonies recovered determined the rate of spontaneous de novo appearance of [PSI+] to be 7.1 × 10−7/generation for a [pin−] strain (Allen et al. 2007) and 5.8 × 10−7/generation in a strain whose [PIN+] status was not mentioned (Lancaster et al. 2010). The frequencies of the spontaneous appearance of [URE3], in a strain with unknown [PIN+] status (Maddelein and Wickner 1999), and of [PIN+] (Sideri et al. 2011) were estimated as 2 × 10−6 and 2 × 10−2, respectively. The high frequency of spontaneous [PIN+] appearance may explain why [PIN+] is found in nature (see below, Biological Effects of Prions).
Prolonged incubation in the cold has been reported to enhance the spontaneous appearance of [URE3] (M. Aigle data, confirmed and cited in Chernoff et al. 1995), [PSI+], and [PIN+] (Derkatch et al. 2000). [PSI+] is also induced by other long-term (rather than short-term) stresses (Tyedmers et al. 2008) that may be related to the accumulation of misfolded proteins and/or alterations of Hsp levels during the stress. Elimination of ribosome-associated peroxiredoxins, Tsa1 and Tsa2, results in the increased induction of [PSI+] by oxidative stress (Sideri et al. 2010) because methionine oxidation in Sup35 apparently promotes aggregation (Sideri et al. 2011). Once again, [PIN+] is required for [PSI+] induction in tsa1Δ tsa2Δ, although formation of [PIN+] can also be induced by oxidative stress (but not by oxidation of Rnq1 itself) in this strain.
Other host effects on prion formation
Aside from the effects of other prion proteins described above, the cellular control of de novo prion formation is not yet well understood. Chaperones that play a crucial role in prion propagation (see above, Requirements for Prion Propagation: Shearing and Segregation) also modulate de novo prion appearance; however, it is not always possible to conclude whether these chaperones act at the stage of initial prion formation per se or influence propagation and/or detection of the newly formed prions. Alterations of the heat-shock factor (Hsf), which regulates Hsp expression, influence de novo [PSI+] appearance. Depending on the Hsf domain altered, these mutations can increase or decrease the frequency of [PSI+] appearance and change the spectrum of the de novo-induced [PSI+] variants (Park et al. 2006). Hsp104 was implicated in the promotion of amyloid formation by excess Sup35NM in vitro (Shorter and Lindquist 2004), although this effect could be due to multiplying the initially formed nuclei via fragmentation. In vivo, excess Hsp104 also promotes de novo induction of the [URE3] prion in the presence (but not in the absence) of the [PIN+] prion, possibly via sheering [PIN+], thereby increasing the abundance of the [PIN+] nuclei (Kryndushkin et al. 2011b). Ssa overproduction increases [PSI+] induction by excess Sup35 (Allen et al. 2005), while deletion of both SSB genes increases both overproduction-induced and spontaneous [PSI+] formation (Chernoff et al. 1999). Therefore, Ssb depletion manifests itself as a “protein mutator,” increasing the frequency of heritable conformational changes in other proteins. As Ssb is implicated in the folding of nascent polypeptides, it may antagonize the accumulation of misfolded protein, providing a substrate for prion nucleation. However, dependence of the effects of Ssa overproduction and Ssb depletion on the presence of a pre-existing nucleus (e.g., [PIN+] prion) indicates that these chaperones do not directly control the nucleation step. Overproduction of Sse1, a nucleotide exchange factor for Ssa, promotes de novo [PSI+] induction, while deletion of SSE1 inhibits it and allows formation of only unstable very weak prion variants (Fan et al. 2007). In contrast, excess Ssa, Ydj1, or Sse1 antagonizes induction of the [URE3] prion (Kryndushkin et al. 2011b). All of these effects are [PIN+]-dependent.
Alterations in the ubiquitin system, which is involved in protein degradation, also influence de novo [PSI+] formation. [PSI+] induction by excess Sup35 is more efficient at increased ubiquitin levels and is reduced by a decrease in the levels of free ubiquitin (i.e., in strains defective for a deubiquitinating enzyme) (Chernova et al. 2003). Deletion of UBC4, which encodes one of the major yeast ubiquitin-conjugating enzymes, increases both [PSI+] resistance to curing via an excess of chaperone Hsp104 and de novo [PSI+] formation (Allen et al. 2007). Notably, the increase of [PSI+] formation by ubc4Δ is independent of the presence of any other prion, although it requires the presence of the Rnq1 protein, even though in a non-prion state. The simplest explanation for the ubc4Δ effect would be that a defect in ubiquitination prevents degradation of misfolded Sup35, thereby increasing its abundance and conversion into a prion. However, no evidence for direct ubiquitination of Sup35 was found. On the other hand, ubc4Δ increases the level of Ssa chaperone associated with Sup35 (Allen et al. 2007). Thus, alterations in the ubiquitin system may influence prions via auxiliary factors.
Several mutations and deletions influencing [PSI+] induction by excess Sup35 have been reported (Bailleul et al. 1999; Ganusova et al. 2006; Tyedmers et al. 2008; Manogaran et al. 2011). Most of these include components of the stress response pathways, ubiquitin-proteasome system, intracellular trafficking networks, and actin cytoskeleton. Mutation in actin or deletions of the genes coding for the actin assembly proteins Sla1, Sla2, or End3 (Bailleul et al. 1999; Ganusova et al. 2006), as well as Las17, Sac6, or Vps5 (Manogaran et al. 2011), decrease both formation of aggregated structures and de novo [PSI+] induction. In contrast, deletions of genes coding for actin assembly proteins Arf1 and Bem1, vesicle-trafficking protein Bug1, and the regulator of osmotic response and actin polymerization Hog1 reduced [PSI+] induction without affecting filament formation, indicating that these proteins probably act at later stages in the pathway (Manogaran et al. 2011). Protein Lsb2/Pin3, containing the QN-rich domain and shown to substitute for [PIN+] when overproduced (Derkatch et al. 2001), is capable of aggregation and [PSI+] induction only when it is associated with the actin cytoskeleton via interactions with Las17 (Chernova et al. 2011). Notably, Lsb2 is a short-lived protein, rapidly induced during heat shock and other stresses, and then degraded via the ubiquitin system. It was shown to be involved in protection of a weak [PSI+] variant from destabilization by heat shock (see Requirements for Prion Propagation: Shearing and Segregation) and may mediate effects of stresses and the ubiquitin system on prion induction (Chernova et al. 2011).
Overall, de novo prion formation, when induced by overproduction of Sup35 or its PrD, appears to be a multi-step process that involves the concentration of misfolded protein in quality control deposits (possibly including IPOD-like structures), followed by subsequent conversion into a prion state. Thus, quality control deposits may act as prion induction sites. Initial aggregate assembly appears to be driven by cytoskeleton-associated proteins, while other proteins may influence prion conversion. Filamentous ring-like agglomerates of misfolded Sup35 are cytotoxic (Zhou et al. 2001; Ganusova et al. 2006; Manogaran et al. 2011); thus, conversion of a misfolded protein into a prion may help to ameliorate toxicity when protein degradation fails to eliminate aggregates. It is possible that a similar pathway is activated when misfolded proteins are accumulated during stress (Chernova et al. 2011). Notably, [PSI+] prions are frequently unstable when they first appear. Over time stable prion variants emerge although some variants remain unstable indefinitely (Derkatch et al. 1996, 2000; J. Sharma and S. W. Liebman, unpublished results).
Sequence Specificity of Prion Transmission and Transmission Barriers
Evolution of prion domains
Yeast PrDs typically evolve faster than regions of the same proteins that are responsible for their major cellular functions. The Sup35C regions in the nearest relatives, Saccharomyces paradoxus and S. cerevisiae (Scannell et al. 2007), are 100% conserved, while the Sup35N and Sup35M regions, respectively, are 94 and 87% conserved (Jensen et al. 2001; Chen et al. 2007). Nonetheless, comparison between these species confirms that the Sup35N sequence remains under selective pressure (Jensen et al. 2001). Sup35 orthologs from genera other than Saccharomyces exhibit an even higher divergence in the Sup35N and M regions, with only 30–40% amino acid identity between S. cerevisiae and distantly related genera such as Pichia or Candida (for reviews, see Inge-Vechtomov et al. 2007; Bruce and Chernoff 2011). At such divergence, only conservation of some patterns of amino acid composition and organization of Sup35NM regions (such as the presence of the QN-rich stretch and oligopeptide repeats in Sup35N and enrichment of Sup35M by charged residues) can be clearly seen. In contrast, the Sup35C regions remain clearly aligned. Despite PrD divergence, most Sup35 proteins of even distantly related yeast species that have been tested are capable of forming a prion in S. cerevisiae. A likely exception is the Sup35 protein of Schizosaccaromyces pombe, which apparently lacks a PrD. Prion formation by Sup35 proteins of non-S. cerevisiae origin in their native environments has not yet been systematically studied, although aggregates of endogenous Sup35 in K. lactis have been reported (Nakayashiki et al. 2001). In the case of Ure2, most proteins of heterologous origin can also form a prion in S. cerevisiae; however, some, e.g., the Ure2 protein of Saccharomyces castellii (Edskes and Wickner 2002; Edskes et al. 2009) or K. lactis (Safadi et al. 2011), were unable to do so. It was also shown that Saccharomyces bayanus Ure2 can form a prion in its native environment while S. paradoxus Ure2 cannot (Talarek et al. 2005). Overall, these results indicate that the ability to form a prion state is generally conserved across long evolutionary distances; however, it can be lost in specific cases. It remains unclear to what extent this ability is realized by the respective proteins in their native proteomes.
Prion species barrier at high levels of sequence divergence
Each particular amyloid fiber typically incorporates only molecules of its specific sequence. The ability of amyloid proteins to form homogenous polymers depends on a high level of sequence identity between the units of a polymer and a newly captured protein molecule. In mammals, even transmission of the prion state to certain homologous proteins from closely related species is inefficient, resulting in the so-called “species barrier” (Moore et al. 2005; Collinge et al. 2006). If a species barrier is overcome, this may lead to cross-species prion transmission, e.g., in the case of “mad cow” disease transmitted to humans.
In yeast, species barriers were initially detected between the S. cerevisiae Sup35 protein and its orthologs from the distantly related species, e.g., Pichia methanolica or C. albicans, whose PrDs show only 30–40% amino acid identity with S. cerevisiae (Chernoff et al. 2000; Kushnirov et al. 2000a; Santoso et al. 2000; Zadorskii et al. 2000). These heterologous proteins do not coaggregate due to divergence of their QN-rich regions (Santoso et al. 2000). Chimeric PrD, composed of portions of the S. cerevisiae and C. albicans QN-rich regions, exhibited a “promiscuous” prion behavior, indicating that each QN-rich fragment works independently (Chien and Weissman 2001; Foo et al. 2011). Heterologous coaggregation with their S. cerevisiae counterpart was reported for the Sup35 orthologs of K. lactis and Yarrowia lipolytica, that are less divergent from Saccharomyces than Candida and Pichia (Nakayashiki et al. 2001). However, it was not clear whether or not coaggregation is followed by transmission of the prion state.
Prion species barrier at low levels of sequence divergence
A prion species barrier was also observed at short phylogenetic distances, e.g., among Ure2 proteins from various species of the genus Saccharomyces (Edskes and Wickner 2002; Baudin-Baillieu et al. 2003; Talarek et al. 2005; Edskes et al. 2009). In these studies, barriers were detected for some but not all species combinations. Different [URE3] prion variants generated by protein with the same sequence could exhibit different cross-species transmission patterns (Edskes et al. 2009).
The most detailed information about the specificity of prion transmission between closely related proteins has been obtained for the Sup35 orthologs from the species of the Saccharomyces sensu stricto group, including S. cerevisiae, S. paradoxus, S. mikata, S. kudriavzevii, and S. bayanus (see Scannell et al. 2007). Sup35 PrDs of the Saccharomyces sensu stricto clade exhibit from 77 to 94% amino acid identity (Jensen et al. 2001; Chen et al. 2007). This is similar to the range of variation observed for vertebrate prion proteins. Both complete Sup35 proteins and chimeric constructs with a heterologous PrD region were studied (Chen et al. 2007, 2010; Afanasieva et al. 2011; see also Bruce and Chernoff 2011). SUP35 genes of different origins (or chimeric genes) were substituted for S. cerevisiae SUP35 by plasmid shuffle in a [PSI+] S. cerevisiae cell. Some data were also confirmed by exposing [psi−] cells to nonhomologous [PSI+] seeds using cytoplasm exchange (cytoduction) in the S. cerevisiae genotypic environment. Coaggregation of proteins containing heterologous PrDs with endogenous Sup35 was detected in vivo, depending on the species combination and [PSI+] variant (Chen et al. 2007, 2010; Afanasieva et al. 2011). However, impairment of cross-species prion transmission was detected even in some combinations where coaggregation was observed. Therefore, the prion species barrier in yeast can be controlled at steps other than the physical association of heterologous proteins. As in the case of Ure2 (see above), the Sup35 species barrier depends not only on sequence divergence, but also on the particular prion variant. Asymmetry of cross-species prion transmission was also detected in some combinations; e.g., prion transfer was inefficient from S. cerevisiae to S. bayanus PrD but efficient in the opposite direction (Chen et al. 2010). Major parameters of the transmission barrier were reproduced in vitro by using purified NM fragments of the S. cerevisiae, S. paradoxus, and S. bayanus Sup35 proteins. With one exception, in vitro results followed in vivo data (Chen et al. 2007).
Experiments with chimeric PrDs have surprisingly shown that different regions of PrD (and not necessarily the QN-rich stretch) are primarily responsible for the species barrier in different combinations (Chen et al. 2010). Moreover, naturally occurring polymorphisms in the non-QN-rich portion of Sup35N or in Sup35M may generate prion transmission barriers even within the S. cerevisiae species (Bateman and Wickner 2012). Notably, transmission barrier is not directly proportional to sequence divergence (Chen et al. 2010; Bruce and Chernoff 2011). These data, in agreement with previous observations in mammalian systems (Prusiner et al. 1990), clearly show that the identity of specific sequences rather than the overall level of PrD homology is crucial for prion transmission.
Transmission barriers generated by mutations
Transmission barriers between yeast prion proteins can also be generated by mutations. Substitutions within Sup35 PrD, e.g., the dominant negative G58D (PNM2 in the oligopeptide repeat region), prevent transmission of some but not other prion variants from wild type to mutant protein (Cox 1994; Doel et al. 1994; DePace et al. 1998; Derkatch et al. 1999; King 2001; Disalvo et al. 2011; Lin et al. 2011). Although the mechanisms are not entirely clear, it is known that the PNM2-dependent transmission barrier is modulated by the Hsp104 chaperone dosage: it becomes more pronounced when the Hsp104 dosage is increased, while decreasing the Hsp104 dosage partly overcomes the barrier (DiSalvo et al. 2011). Apparently sequence differences generate metastable heteroaggregates that are more sensitive to the disaggregating activity of Hsp104. A similar mechanism was proposed to explain the decreased conversion of heteroaggregates into stably maintained prions for combinations of the divergent Saccharomyces sensu stricto Sup35 proteins (Afanasieva et al. 2011). PNM2-containing heteroaggregates also show asymmetric distribution to mother cells in cell divisions (Verges et al. 2011). This could also be explained by differential sensitivity to Hsp104. Another dominant negative mutant, in the QN-rich region of the Sup35 PrD, Q24R, reduces the ability of the protein to be converted to the prion by wild-type Sup35 aggregates (DiSalvo et al. 2011).
For the Rnq1 protein, transmission barriers were shown to be generated by deletions of QN-rich regions (Kadnar et al. 2010). Thus, transmission specificity is controlled not only by the identity of amino acid sequences, but also by the length of cross-interacting regions. This agrees with previous observations that shortened PrD fragments can “poison” propagation of the [URE3] prion (Edskes et al. 1999).
Fidelity of cross-species prion conversion
Rare instances of prion transmission to highly divergent PrDs, e.g., from P. methanolica to S. cerevisiae, resulted in multiple prion variants (Vishveshwara and Liebman 2009). This could be explained by rare nucleation of de novo amyloid formation by the aggregate of a divergent protein, similar to cross-seeding by nonhomologous proteins with PrDs of similar amino acid compositions (see Spontaneous and environmentally induced prion formation).
Transmission of the prion state between Sup35 or Ure2 proteins with closely related or mutationally altered PrDs sometimes resulted in prions with altered phenotypic patterns. Notably, reverse transmission of the prion state back to the original protein restored the original prion patterns in most (Edskes et al. 2009; Chen et al. 2010; Lin et al. 2011) but not all (e.g., S. cerevisiae/S. bayanus) cases (Chen et al. 2010). As the frequency of reverse prion transmission in this combination was reasonably high, the appearance of new prion variants could not be explained simply by nonspecific de novo nucleation. The data suggest that conformational fidelity during prion transmission is controlled at both genetic (sequence identity) and epigenetic (prion variants) levels. Notably, the OR region controls both the transmission barrier and the conformational infidelity in the S. cerevisiae/S. bayanus combination.
Overall, the data suggest that both specificity and conformational fidelity of prion transmission in yeast are determined by relatively short amino acid stretches, rather than by the whole PrD (Chernoff 2008; Bruce and Chernoff 2011). These stretches could correspond to sequences that initiate intermolecular interactions, resulting in the formation of the cross-β amyloid core. Indeed, stronger [PSI+] variants need shorter portions of PrD for their propagation, both in vivo (Shkundina et al. 2006) and in vitro (Chang et al. 2008). The locations of the specificity stretches could control the variant-specific prion patterns by determining the position and size of the amyloid core(s). In the case of a variant switch (e.g., the S. cerevisiae/S. bayanus combination), the interaction between altered specificity stretches is probably too weak to form the cross-β structure. However, alternative stretches located at different positions could occasionally be employed, resulting in the formation of an amyloid core at a different location and/or size. In such a scenario, the ability of prion variants either to be faithfully propagated by a heterologous protein or to undergo a conformational switch depends on the level of identity of the specificity stretches that define amyloid cores in these variants (Figure 12).
Antagonism between heterologous prions
Although, as described in the previous sections, heterologous prions enhance each other’s appearance, the newly established prion does not require the presence of the heterologous prion once it has been established (Derkatch et al. 2000). Rather, it is often the case that a prion will destabilize a heterologous prion. For example, certain variants of [PIN+] destabilize weak [PSI+] (Bradley and Liebman 2003). While the mechanism of destabilization is unknown, [PIN+] causes [PSI+] aggregates to become larger, and enlarged aggregates are harder to transmit to daughters (Mathur et al. 2009). Also, [PSI+] and [URE3] destabilize each other (Bradley et al. 2002; Schwimmer and Masison 2002). Overexpression of fragments of the Ure2 prion domain cure [URE3] (Edskes et al. 1999), and overexpression of certain fragments of the Rnq1 prion domain in the presence of [PIN+] inhibits propagation of [PSI+] and [URE3] (Kurahashi et al. 2008). Finally, overexpression of Rnq1with certain mutations in the non-prion domain causes [PSI+] aggregates to enlarge and be lost (Kurahashi et al. 2011). Indeed, overexpression of a large number of QN-rich proteins destabilize pre-existing [PSI+] and [URE3] prions apparently by interfering with the ability of the prion aggregates to be sheared (Z. Yang, J. Hong, I. L. Derkatch, and S. W. Liebman, unpublished results).
Prions of normal yeast proteins
In the past few years, it has become clear that prion phenomena are much more widespread than previously thought. Seven currently proven amyloid-based yeast prions and their properties are summarized on Table 1. All have QN-rich domains, although this is biased by the screening criteria used (see below). Curiously, four of the six prions with known functions are involved in transcriptional regulation. The prion aggregates generally appear in the cytoplasm, but one prion, [ISP+], exhibits nuclear aggregates (Rogoza et al. 2010). [PSI+] and [URE3] were initially identified as prions on the basis of their unusual genetic properties (Wickner 1994). Later, [PIN+]/[RNQ+] was identified as a prion (see Heterologous prion cross-seeding). Other proteins were identified as possible prion candidates on the basis of sequence algorithms derived from known QN-rich prion domains (Michelitsch and Weissman 2000; Santoso et al. 2000). Some of the QN-rich domains from these proteins are able to confer a prion state to a chimeric reporter construct. A recent large-scale screen (Alberti et al. 2009) identified 19 new potential PrDs that are capable of replicating as a prion in cells when fused to a reporter. It is likely that many of these domains can also confer a prion state to their native proteins, although an example exists of a protein (New1) that contains a PrD region conferring a prion state to a reporter but does not appear to form a prion on its own (Santoso et al. 2000; Osherovich et al. 2004). Indeed, undoubtedly, there are more prions because the screen was not exhaustive, the initial sequence criteria were biased (so that one independently identified prion protein, Sfp1, was excluded on this basis), and some prion proteins (Cyc8 and Mot3) did not work with the Sup35MC-based reporter used.
An earlier screen for proteins that, when overexpressed, substitute for [PIN+] in promoting the appearance of [PSI+], identified 11 candidates (Derkatch et al. 2001) with QN-rich domains, including Ure2 and the above-mentioned New1, also proven to possess such a property in an independent article (Osherovich and Weissman 2001). The Alberti et al. (2009) candidates overlap with some of these proteins, and two of them, the chromatin remodeler Swi1 and the global transcriptional regulator Cyc8, have now been shown to, respectively, form the [SWI+] (Du et al. 2008) and [OCT+] (Patel et al. 2009) prions. However, other proteins from this screen, e.g., Pin3/Lsb2, do not appear to propagate as a typical stable prion (Chernova et al. 2011).
The [ISP+] prion was first detected because it counteracted the nonsense suppression associated with certain mutations (Aksenova et al. 2006). It was later identified as a prion form of the global transcriptional regulator, Sfp1, which has a QN-rich domain (Rogoza et al. 2010). Antisuppression was probably due to increased abundance of a reporter transcript; in contrast to five other proven prion proteins with known functions, the [ISP+] phenotype is different from a loss of Sfp1 function. [ISP+] is transferred by cytoduction less efficiently than other prions, probably due to the nuclear location of [ISP+] prion aggregates. Also, propagation of [ISP+] also does not require Hsp104, which is needed for cytoplasmic prions (see Requirements for Prion Propagation: Shearing and Segregation).
In addition to the amyloid-based prions, other transmissible protein-based phenomena have been described in yeast. The [GAR+] prion is propagated by a complex between two proteins, Std1 (a glucose-signaling protein) and Pma1 (a plasma membrane proton pump). This atypical prion causes growth on glycerol in the presence of a non-metabolizable glucose analog as a dominant non-Mendelian trait. Propagation of [GAR+] is not dependent upon Hsp104, and the Std1-Pma1 prion complex does not show the characteristics of amyloid (Brown and Lindquist 2009).
Another transmissible protein-based phenomenon is based on self-catalyzed protein processing rather than on protein aggregation. The protein involved is protease B, which can spontaneously generate a persisting self-activated state, termed [β], in yeast cells that are missing the functionally redundant protease A protein (Roberts and Wickner 2003).
Other non-Mendelian elements of a yet-unknown nature, but with prion-like behavior, have also been reported in yeast. One example is [NSI+], a non-Mendelian nonsense suppressor that is detectable only in the absence of the Sup35N prion domain. [NSI+] possesses the genetic characteristics of a classical yeast prion, although its protein determinant is unknown (Saifitdinova et al. 2010).
Artificial and heterologous prions that can propagate in yeast
Several studies have employed fusions of known or suspected PrDs to GFP, Sup35C, Ure2C, or glucocorticoid receptor reporters (Patino et al. 1996; Li and Lindquist 2000; Sondheimer and Lindquist 2000; Alberti et al. 2009, 2010). These chimera often, but not always, behave as prions. Some aggregation-prone sequences, e.g., polyQ stretches of non-yeast origin (Osherovich et al. 2004) or oligopetide repeats of the mammalian prion protein PrP (Parham et al. 2001; Dong et al. 2007; Tank et al. 2007), can substitute for all (if expanded) or part of Sup35 PrD without disrupting its prion properties. Interestingly, a non-QN-rich PrD of the Podospora Het-s protein, fused with GFP, was shown to propagate as a prion in yeast, making it likely that non-QN-rich endogenous yeast prion proteins also exist (Taneja et al. 2007). Indeed, see Suzuki et al. (2012), published while this review was in press.
Biological Effects of Prions
Amyloids and amyloid-like inclusions are associated with diseases in humans and other mammals, such as Alzheimer’s, Parkinson’s, and Huntington’s diseases; amyotrophic lateral sclerosis; type II diabetes; transmissible spongiform encephalopathies (“mad cow” disease, Creutzfeldt-Jacob disease, etc.); and others (Aguzzi and O’Connor 2010). In yeast, the presence of the [URE3] prion, or a combination of the [PSI+] prion with the tRNA suppressor SUQ5 (Eaglestone et al. 1999; Jung et al. 2000; Schwimmer and Masison 2002), can induce the stress response. Overproduction of Sup35 or its PrD is toxic to [PSI+] strains (Chernoff et al. 1992) and at high levels to [PIN+] strains, in which de novo [PSI+] induction is efficient (Derkatch et al. 1997), but not to strains lacking any prions. Likewise, Rnq1 overproduction is toxic to [PIN+] strains (Douglas et al. 2008). Overproduced Sup35 PrD in a [PSI+] cell sequesters full-length Sup35 into prion aggregates, and overproduced full-length Sup35 sequesters another release factor, Sup45, contributing to toxicity (Vishveshwara and Liebman 2009). Indeed, some sup45 mutants (Kiktev et al. 2007, 2011) or a heterozygous sup45 deletion (Dagkesamanskaya and Ter-Avanesyan 1991) are lethal or sublethal in a [PSI+] background. Although some variants of [PSI+] do not impact exponential yeast growth (Eaglestone et al. 1999), other [PSI+] variants are toxic unless rescued by the Sup35 derivative (Sup35C) that lacks the PrD and therefore cannot be sequestered (McGlinchey et al. 2011). Some variants of the [URE3] prion also decrease growth (McGlinchey et al. 2011). Some Hsp104 mutations result in [PSI+]-dependent cytotoxicity (Gokhale et al. 2005). Overall, existing evidence indicates that at least some [PSI+] and [URE3] variants are detrimental to yeast (see Wickner et al. 2011). It was hypothesized (Bateman and Wickner 2012) that sequence polymorphisms causing prion transmission barriers arose to prevent acquisition of harmful prions.
Prions as susceptibility factors for polyQ disorders
Prion variants that do not cause toxicity on their own may become toxic in combination with other factors. For example, fragments of the human huntingtin protein with the expanded polyQ stretch, associated with Huntington’s disease, are toxic to yeast strains containing an endogenous prion, such as [PIN+] (Meriin et al. 2002) and/or [PSI+] (Gokhale et al. 2005). Heterologous pre-existing aggregates promote polyQ aggregation in yeast (Osherovich and Weissman 2001; Meriin et al. 2002), and [PIN+] apparently mediates sequestration of some actin assembly proteins (Meriin et al. 2003), while [PSI+] mediates sequestration of Sup45 (Gong et al. 2012; Kochneva-Pervukhova et al. 2012) in the presence of polyQ aggregates. This leads to a cytotoxic defect, establishing yeast as a model for Huntington’s disease (see Duennwald 2011; Mason and Giorgini 2011) and raising the intriguing possibility that variations in the onset of Huntington’s disease in humans, recorded even among individuals with the same length of polyQ stretch, could be partly explained by variations in the composition of endogenous QN-rich aggregates in human cells.
Facts and hypotheses about biologically positive roles of prions in yeast and other fungi
The finding that most proteins can form amyloids in vitro suggests that the amyloid conformation is an ancient protein fold (Chiti and Dobson 2009). Indeed, positive biological roles of amyloids include scaffolding of melanin polymerization (Fowler et al. 2007), storage of peptide hormones (Maji et al. 2009), protection from stress (Iconomidou and Hamodrakas 2008), silk production (Romer and Scheibel 2008), substrate attachment (Gebbink et al. 2005), biofilm formation (Wang et al. 2008), and a proposed connection to long-term memory (Si et al. 2003, 2010; Heinrich and Lindquist 2011). [Het-s] Podospora was the first prion shown to provide a biological advantage to its host (Coustou et al. 1997; Wickner 1997). [Het-s] controls vegetative incompatibility, an adaptive trait, by causing death of [Het-S] mycellium at the position of contact (see Saupe 2007, 2011). Initially, only [PIN+] and not [PSI+] and [URE3] was found in any natural or industrial Saccharomyces strains (Chernoff et al. 2000; Resende et al. 2003; Nakayashiki et al. 2005). However, a more extensive search recently identified [PSI+] in ∼1.5% of strains (Halfmann et al. 2012). Notably, all of these [PSI+] strains were also [PIN+], suggesting that [PIN+] promotes [PSI+] generation in the wild as well as in the laboratory. If transcriptional regulators form prions, that could potentially result in regulatory switches that are adaptive. One intriguing candidate is Mot3, which regulates cell-wall biosynthesis genes, shows a high frequency of spontaneous conversion between prion and non-prion states (Alberti et al. 2009), and is found in a significant fraction of wild strains (Halfmann et al. 2012). About one-third of the natural and industrial Saccharomyces strains tested exhibit phenotypes that are curable by transient inactivation of Hsp104 (Halfmann et al. 2012), indicating that yet-unidentified Hsp104-dependent prions are widespread in wild yeast and may contribute to adaptation. Readthrough of termination codons caused by [PSI+] was proposed to broaden the adaptation spectrum in changing environments via increasing phenotypic variability (True and Lindquist 2000; True et al. 2004). This agrees with the hypothetical consequences of translational ambiguity in general (e.g., see Inge-Vechtomov et al. 1994) and is supported by mathematical simulations (Masel and Bergman 2003; Masel 2006; Griswold and Masel 2009). [PSI+] indeed increases the growth of some laboratory or wild strains in certain conditions (True and Lindquist 2000; True et al. 2004; Halfmann et al. 2012), although these effects are genotype-specific, and no consistent effect of [PSI+] on adaptation to novel environments has been detected (Joseph and Kirkpatrick 2008). Some [PSI+]-mediated phenotypes could be due to suppression of nonsense mutations occurring in the genomes of the respective strains and/or due to induction of stress-defense genes (Eaglestone et al. 1999; Schwimmer and Masison 2002). It is not known to what extent [PSI+] influences termination at natural stop signals that are protected by context (Bonetti et al. 1995). However, [PSI+] modulates at least one naturally occurring mistranslation phenomenon, frameshifting in the antizyme gene, responsible for feedback regulation of polyamine biosynthesis (Namy et al. 2008). This explains some of the [PSI+]-mediated phenotypes, although their adaptive role remains unclear (see Chernoff 2008).
Notably, pathogenic and beneficial effects of a prion are not mutually exclusive. Considering prion formation as a mutation occurring at the protein level, with prion variants analogous to alleles (Chernoff 2001), one could expect different effects on fitness for different prion variants, in the same way as mutations with deleterious or beneficial effects may arise in a single gene. Environmental changes may affect selection in the following way: if normal function of a prion protein becomes deleterious under certain conditions, decrease of this function in a prion form would become adaptive. Thus, prions could be helpful in certain conditions and be eliminated otherwise.
It is also possible that, in some cases, not the prion per se but rather the process leading to its formation could be beneficial. For example, aggregates formed during unfavorable conditions may be beneficial due to their increased resistance to degradation (see Chernoff 2007). While such aggregates are normally disassembled by chaperones upon return to normal conditions, prions could arise as by-products of such a process. Indeed, stress granules, assemblies that protect pre-initiation mRNA complexes during stress, are formed with participation of a protein containing a prion-like QN-rich domain (Gilks et al. 2004). Filamentous aggregates (rings) of overproduced Sup35 protein are toxic (see De Novo Prion Formation), while prion formation may ameliorate this toxicity, becoming a “lesser of two evils” (Ganusova et al. 2006). Possibly, similar events may occur if misfolded protein is accumulated during stress.
While the complete spectrum of biological effects of prions is yet to be uncovered, a growing number of prion examples indicate that the impact of prions on the host biology is significant and can no longer be ignored in models of evolution.
At the time at which the precursor of this YeastBook series (the second three-volume Cold Spring Harbor monograph series entitled The Molecular and Cellular Biology of Yeast Saccharomyces) was published in 1992, the basis of the mysterious [URE3] and [PSI+] non-Mendelian factors was unknown. Today, their identification as prions has established the entirely new field of protein-based inheritance. Along with other epigenetic phenomena, prions have made it increasingly clear that knowing the sequence of the genome does not tell the whole story. Thus, a key remaining issue is the biological role of prions, including the contribution of protein-based components toward the transmission of heritable traits. Yeast will continue to lead the way toward understanding this and other fundamental properties of prions. By the time the next encyclopedia of yeast biology is published we will hopefully know how prion seeds are dispersed during cell division, what the structural basis for prion variants is, and what rules govern the specificity of their reproduction. Since amyloids and prions are found in many organisms, the importance of these findings will extend well beyond yeast and may help to develop therapeutic and prophylactic treatments for human protein-misfolding diseases.
S.W.L. thanks C. V. Vishveshwara for the policemen concept and drawings; F. Arslan, K. Facemyer, and V. Mathur for comments on the manuscript; and K. Facemyer for an introduction to Sketchup. Y.O.C. thanks G. Newnam for technical help with preparation of the manuscript; A. Romanyuk for discussing the model shown in Figure 4; and K. Bruce for help with Figure 12. S.W.L. acknowledges the National Institutes of Health (NIH) (grant R01GM 056350) for financial support. Y.O.C. acknowledges the NIH (grant R01GM 058763) and the National Science Foundation (grant MCB-1024854) for financial support.
Communicating editor: J. Boeke
- Received December 12, 2011.
- Accepted February 20, 2012.
- Copyright © 2012 by the Genetics Society of America