Specificity of Prion Assembly in Vivo

The yeast prions [PSI+] and [PIN+] are self-propagating amyloid aggregates of the Gln/Asn-rich proteins Sup35p and Rnq1p, respectively. Like the mammalian PrP prion “strains,” [PSI+] and [PIN+] exist in different conformations called variants. Here, [PSI+] and [PIN+] variants were used to model in vivo interactions between co-existing heterologous amyloid aggregates. Two levels of structural organization, like those previously described for [PSI+], were demonstrated for [PIN+]. In cells with both [PSI+] and [PIN+] the two prions formed separate structures at both levels. Also, the destabilization of [PSI+] by certain [PIN+] variants was shown not to involve alterations in the [PSI+] prion size. Finally, when two variants of the same prion that have aggregates with distinct biochemical characteristics were combined in a single cell, only one aggregate type was propagated. These studies demonstrate the intracellular organization of yeast prions and provide insight into the principles of in vivo amyloid assembly.

The formation of ␤-sheet-rich amyloid aggregates is a hallmark of many neurodegenerative disorders in mammals (1). Deposits of the self-propagating infectious prion protein PrP in the brain cause severe neuronal damage associated with transmissible spongiform encephalopathies (2)(3)(4). Expansion of polyglutamine stretches in huntingtin and MJD proteins leads to their aggregation as amyloids, which accompany Huntington's and Machado-Joseph diseases, respectively (5). In Saccharomyces cerevisiae stably inherited epigenetic factors [PSI ϩ ] (6) and [PIN ϩ ] (7,8) were established to be the prion forms of glutamine-and asparagine-rich (Gln/Asn-rich) proteins Sup35p and Rnq1p (9 -11). [PSI ϩ ] and [PIN ϩ ] provide invaluable tools in the investigation of the principles of organization and interactions among amyloid aggregates in vivo.
Sup35p is an essential protein, and in the non-prionized soluble form, referred to as [psi Ϫ ], it is involved in the termination of protein translation. In [PSI ϩ ] cells the pool of active Sup35p is severely depleted causing reduced translational termination efficiency (for reviews see Refs. [12][13][14]. Intracellularly, [PSI ϩ ] is assembled into large aggregates containing SDS-stable Sup35p polymers (made of 9 -50 monomers) and possibly other proteins (15). RNQ1 is not essential (16), and the role of the non-prionized Rnq1p, referred to as [pin Ϫ ], is un-known. When in the [PIN ϩ ] state, Rnq1p dramatically enhances the rate of [PSI ϩ ] appearance, possibly by templating the prion conformation of Sup35p (11,17).
Deletion of Hsp104p, a yeast chaperone involved in the refolding of aggregated or misfolded proteins, leads to loss of [PSI ϩ ] and [PIN ϩ ] (7,10,18,19). Stable inheritance of the yeast prions requires each cell to contain discrete prion aggregates that can migrate to daughter cells during budding (20 -22). Strong evidence supports the idea that Hsp104p maintains a sufficient number of prion aggregates by shearing them, thus securing successful prion transmission in the growing culture. Guanidine hydrochloride in non-denaturing concentrations blocks Hsp104p activity, which leads to an uncontrolled growth of the Sup35p polymers followed by prion loss (15,20,(22)(23)(24)(25). Interestingly, although guanidine often causes the simultaneous loss of [PSI ϩ ] and [PIN ϩ ] in two-prion [PSI ϩ ] [PIN ϩ ] cells, they can also be lost independently (7). This independence raises the question of whether two-prion cells contain homogeneous or heterogeneous prion bodies.
Like human amyloidogenic proteins, Sup35p and Rnq1p form self-seeding amyloid fibers in vitro (10, 26 -29) and in vivo (30). It has also been shown recently that preformed Rnq1p fibers facilitate polymerization of the Gln/Asn-rich domain of Sup35p (17). Using fluorescently labeled Sup35p and Rnq1p, newly forming Sup35p aggregates that appeared during [PSI ϩ ] induction were found to co-localize with Rnq1p aggregates. However, once [PSI ϩ ] was established, the fluorescent aggregates co-localized less frequently (17). These observations are consistent with the hypothesis that [PIN ϩ ] aggregates interact with Sup35p and template [PSI ϩ ] induction. Interestingly, a number of other Gln/Asn-rich proteins can substitute for [PIN ϩ ], suggesting the importance of Gln/ Asn-rich sequences in the physiology of amyloid formation (11,31).
The [PSI ϩ ] and [PIN ϩ ] prions (19,32) as well as the mammalian PrP prion (reviewed in Ref. 33), can exist as different "strains," also called variants, determined by a specific conformation of the prion molecules (34 -37). Weak [PSI ϩ ] variants are generally less mitotically stable than strong [PSI ϩ ] variants (19) and have more non-prionized (soluble) Sup35p molecules in the cytoplasm (29,38). In a cross of strong and weak [PSI ϩ ] haploids the diploids and their progeny exhibit the strong [PSI ϩ ] phenotype, implying that strong [PSI ϩ ] variants are "dominant" over weak (19,38). It is unclear, however, whether the prion variant dominance phenomenon is confined to a phenotypic dominance only (i.e. aggregates of both variants co-exist) or is accompanied by a complete elimination of the "recessive" variant aggregates.
[PIN ϩ ] variants have been distinguished by their fluorescent pattern of Rnq1p-green fluorescent protein fusion staining (mostly single large fluorescent dots (s.d.) 1 versus multiple small (m.d.) fluorescent dots per cell) (39) and by the efficiency with which they facilitate the induction of [PSI ϩ ], ranging from low to very high (32). Variant dominance was shown to correlate inversely with the amount of non-prionized Rnq1p (32). Therefore it appears that the dominance of a prion variant is determined by the efficiency with which its aggregates recruit new molecules (32,39). This is supported by the finding that protein extracts from cells carrying strong [PSI ϩ ] variants were capable of converting non-prionized Sup35p into fibers more efficiently than the extracts from weak [PSI ϩ ] (29,40 (39). These strains were crossed to L1767 ⌿ Ϫ. The diploids resulting from these crosses were selected on the media deficient in adenine.
Preparation of Yeast Cell Lysates-Yeast cultures were grown in liquid YPD media to A 600 of 1.5-2.0. For investigations of the effect of guanidine hydrochloride, cells were grown in YPD to A 600 of 1.5-2.0, and then cultures were divided in half, and the initial volume was reconstituted with YPD. Guanidine HCl was added to 7 mM final concentration to one of the cultures, and cells were grown until the A 600 of the diluted cultures doubled. The cells were harvested, washed with water, resuspended, and lysed by vortexing with 0.5-mm glass beads in a protein extraction buffer containing antiproteases (PEB/AP: 25 mM Tris-HCl (pH 7.5), 50 mM KCl, 10 mM MgCl 2 , 1 mM EDTA-Na, 5% glycerol, 100 M 1-chloro-3-tosylamido-7-amino-2-heptanone (TLCK), 100 M L-1-tosylamido-2-phenylethyl chloromethyl ketone (TPCK), 10 mM 1,10-phenanthroline, 94 M leupeptin, 1 M pepstatin A, 5 g/ml aprotinin, 2 mM phenylmethylsulfonyl fluoride, and "protease inhibitor mixture for yeast" (Sigma, 1:50)). Crude lysates were precleared by 1 min of centrifugation at 600 ϫ g at 4°C. Such a short preclearing step allows for efficient preservation of prion aggregates that may pellet along with the cell debris upon higher speed or longer duration of spin. Protein concentration was determined as described (42) using the Bio-Rad protein assay.
Analysis of Yeast Cell Lysates-For sucrose gradient analyses, fresh lysates with or without the addition of SDS to 2% were incubated for 7 min at room temperature and fractionated at 4°C in a swinging bucket rotor through a 20 -60% continuous sucrose/PEB gradient for 40 min at 10,600 ϫ g, unless indicated otherwise. Equal fractions were collected from top to bottom, diluted 1:2 in PEB/AP, resolved in a 10% polyacrylamide gel as described (43), and transferred to an immunoblot polyvinylidene difluoride membrane (Bio-Rad). Sup35p was detected using BE4 monoclonal antibodies against the Sup35p C-terminal domain. Rnq1p was detected by polyclonal antibodies (a kind gift from S. Lindquist). Signal was revealed using a Tropix kit (Applied Biosystems) as suggested by the manufacturer. To stain the membrane with different antibodies, it was stripped twice by incubation for 30 min in 0.2 M glycine (pH 2.2) containing 1% SDS and 1% Tween 20. Using this procedure we obtained the same images regardless of which antibodies were used first.
Semi-denaturing detergent agarose gel electrophoresis (SDD-AGE) was performed as described elsewhere (15) with the following changes. Lysates were incubated for 7 min in sample buffer (60 mM Tris-HCl (pH 6.8), 5% glycerol, 2% SDS, 0.05% bromphenol blue) at the indicated temperatures, resolved in a horizontal 1.5% agarose gel in a standard Tris/glycine/SDS buffer, transferred electrophoretically to a polyvinylidene difluoride membrane, and probed with antibodies as described above. An increase of the final SDS concentration in the sample to 7.5% did not affect the mobility of prion subparticles (not shown). A preparation of chicken pectoralis extract (a kind gift from T. Keller) was used to estimate molecular weight (44). When stained with Coomassie, chicken pectoralis extract reveals several abundant muscular proteins: titin (ϳ3000 kDa), nebulin (ϳ750 kDa), and myosin heavy chain (ϳ200 kDa). Although this ladder cannot be used for precise determination of molecular mass, it does provide an estimate.
Comparison of the levels of soluble Rnq1p was performed as described elsewhere (32) with the following changes. Lysates were obtained as mentioned above, equalized by protein to 4 mg/ml with PEB, and spun for 10 min at 10,000 ϫ g for additional preclearing. Supernatants were subjected to ultracentrifugation at 280,000 ϫ g for 30 min at 4°C. Equal amounts of supernatants were compared by Western analysis as described above. Equal amounts of protein were loaded according to anti-Sup35p and Coomassie staining.  2A). We refer to these SDS-stable structures as prion subparticles. Interestingly, the size distributions of [PIN ϩ ] subparticles were prion variant-specific. Although m.d. [PIN ϩ ] subparticles were evenly distributed and had a size range from ϳ600 to Ͼ3000 kDa, the size of high s.d. [PIN ϩ ] subparticles varied from ϳ400 to Ͻ3000 kDa. After probing the same gel with anti-Sup35p antibodies, we found Sup35p migrated as a monomer, indicating that most of the non-prionized Sup35p does not form stable interactions with subparticles from these [PIN ϩ ] variants ( Fig. 2A).

Rnq1p Is Assembled into Heavy SDS-sensitive Aggregates in [PIN
The size distribution of subparticles from different s.d. [PIN ϩ ] variants that were previously characterized as possessing different [PSI ϩ ] induction strengths (low, medium, and very high) (32) are shown in Fig. 3A (lanes 1, 2, and 4). Subparticles from low, medium, and very high s.d. [PIN ϩ ] had size distributions similar to each other (but not exactly the same) and to m.d. [PIN ϩ ] (Fig. 3B) (Fig. 2B).
Several findings suggest that differences between prion variants are caused by the conformation that the prion molecule acquires (36,37). If this were true then [PIN ϩ ] prion subparticles from different prion variants should have distinct physical properties. Indeed, although m.d. and s.d.
[PIN ϩ ] subparticles were equally stable in 2 and 7.5% SDS (not shown) at room temperature and at 30°C, when subjected to 60°C their stability differed. Although all s.d.
[PIN ϩ ] subparticles broke down into smaller species migrating as Rnq1p monomers (ϳ46 kDa) and at ϳ200 and ϳ400 kDa (Fig. 3B)  When each of the [PIN ϩ ] variants was cytoduced (a process that involves cytoplasmic but not nuclear transfer) into a genetically distinct strain, neither the original size distribution of the [PIN ϩ ] subparticles nor the peculiar thermal susceptibility of m.d. [PIN ϩ ] was altered (compare Fig. 3A, lanes 1, 2, and 4 with Fig. 4A, lanes 2, 4, (Fig. 6). As a control to show that we could detect some size shifts, we cultivated a weak [PSI ϩ ] ⌬rnq1 strain in YPD containing 7 mM guanidine for one generation. Consistent with the results reported previously (15) we observed a significant increase in the size of the Sup35p subparticles (Fig. 6) (Fig. 5A).
Sedimentation Profiles of Prion Aggregates Are Prion Variant-specific-We found that the prion variant-specific sedimentation profiles were different between m.d. and s.d. [PIN ϩ ] and between strong and weak [PSI ϩ ] aggregates (Fig. 7)  files. Weak [PSI ϩ ] aggregates formed a "peak" in the third fraction; strong [PSI ϩ ] aggregates were distributed evenly throughout the gradient (Fig. 7).
Aggregates  (Fig. 7B, and not shown). DISCUSSION To dissect the intracellular organization of [PIN ϩ ] we used SDS treatment to disrupt non-amyloid interactions within the [PIN ϩ ] prion aggregates. SDS-stable subparticles that result from aggregate disassembly were reliably resolved in agarose gels (15), a technique that may prove useful in the investigation of amyloid aggregates and other SDS-stable heavy protein complexes. The two-level structural organization appears to be a general feature of yeast prions, having now been demonstrated for two prions, [PSI ϩ ] (15) and [PIN ϩ ]. Interestingly, the [Hets] prion of the fungus Podospora anserina was shown to be able to form elongated aggregates in vivo that grow by lateral association of shorter fibrillar aggregates (45). The finding that the subparticle size of both [PSI ϩ ] (15) and [PIN ϩ ] increases immediately in response to growth in guanidine supports the current theory that guanidine inhibits the function of Hsp104p as a prion disaggregase, which leads to abrogation of prion aggregate generation via shearing (20,22). We showed here that although all [PIN ϩ ] variants possessed SDS-stable subparticles, they were not uniform in size and thermal stability. Overall, the size distribution of [PIN ϩ ] subparticles was prion variant-specific and corresponded to ϳ20 -100 monomers of Rnq1p. This number is comparable with the 9 -50 monomers of Sup35p reported for [PSI ϩ ] subparticles (15). Importantly, the specific size distribution of the [PIN ϩ ] subparticles from different variants and their susceptibility to heat treatment in the presence of SDS were cytoducible and hence not dependent on the genetic background but rather reflected variations in the physical properties of the prionized Rnq1p. Lysates of 74D-694 derivatives were analyzed as in Fig. 3 ] were cultivated with (ϩ) or without (Ϫ) guanidine (Gu) (see "Experimental Procedures") and analyzed for Sup35p by SDD-AGE as described in Fig. 3.
The size of the subparticles depends on the balance between the processes of new molecule recruitment on the one hand and Hsp104p-dependent subparticle shearing on the other. The essential factor in these processes is the conformation of the prion variants that probably determines the speed of recruitment and the affinity to Hsp104p. The outcome of the interplay between these processes should determine the overall efficacy with which the non-prionized molecules get recruited into the subparticles. Importantly, this efficacy should not necessarily correlate with the subparticle size, for high speed of recruitment may be compensated by low efficiency of shearing by Hsp104p, and vice versa. The low level of soluble protein is therefore a much more informative indicator of a "dominant" variant. The system we described precludes two prion variants with different conformational characteristics from co-existing, simply because one variant will prevail in a competition for the non-prionized molecules, and eventually it will exile the less effective variant from the culture. This is exactly what we found for the [ Even at the level of aggregates [PSI ϩ ] and [PIN ϩ ] remained structurally separated and did not form appreciably tight in-teractions with each other. This does not preclude the possibility that they co-localize in the cell. Neither can we rule out the hypothesis that these aggregates contain other, unidentified proteins. At the molecular level, two models of prion aggregate organization can be suggested. The first model postulates that a direct, SDS-sensitive interaction between subparticles is sufficient to assemble them into mature aggregates, which may or may not be decorated by additional proteins. The second model requires additional proteins to mediate aggregate assembly by forming SDS-sensitive interactions between the subparticles. Whichever model is correct, identification of the attendant proteins may shed light on the biological role of yeast prions and prion variants as well as provide insight into the mechanisms of protein sequestration by toxic prion aggregates that accompany neurodegenerative diseases in mammals (46).
We reported previously (39)  Here we have presented biochemical evidence suggesting that heterologous prions [PSI ϩ ] and [PIN ϩ ] do not form "mixed" subparticles or aggregates. Intracellularly, these prions are highly selective and form rather separate structures, which may be in close proximity to each other but are devoid of tight interactions. Thus, the results presented in this study provided a deeper look into the structural organization of prion particles inside living cells.