Mammalian amyloidogenic proteins promote prion nucleation in yeast

Fibrous cross-β aggregates (amyloids) and their transmissible forms (prions) cause diseases in mammals (including humans) and control heritable traits in yeast. Initial nucleation of a yeast prion by transiently overproduced prion-forming protein or its (typically, QN-rich) prion domain is efficient only in the presence of another aggregated (in most cases, QN-rich) protein. Here, we demonstrate that a fusion of the prion domain of yeast protein Sup35 to some non-QN–rich mammalian proteins, associated with amyloid diseases, promotes nucleation of Sup35 prions in the absence of pre-existing aggregates. In contrast, both a fusion of the Sup35 prion domain to a multimeric non-amyloidogenic protein and the expression of a mammalian amyloidogenic protein that is not fused to the Sup35 prion domain failed to promote prion nucleation, further indicating that physical linkage of a mammalian amyloidogenic protein to the prion domain of a yeast protein is required for the nucleation of a yeast prion. Biochemical and cytological approaches confirmed the nucleation of protein aggregates in the yeast cell. Sequence alterations antagonizing or enhancing amyloidogenicity of human amyloid-β (associated with Alzheimer's disease) and mouse prion protein (associated with prion diseases), respectively, antagonized or enhanced nucleation of a yeast prion by these proteins. The yeast-based prion nucleation assay, developed in our work, can be employed for mutational dissection of amyloidogenic proteins. We anticipate that it will aid in the identification of chemicals that influence initial amyloid nucleation and in searching for new amyloidogenic proteins in a variety of proteomes.


Fibrous cross-␤ aggregates (amyloids) and their transmissible forms (prions) cause diseases in mammals (including humans) and control heritable traits in yeast. Initial nucleation of a yeast prion by transiently overproduced prion-forming protein or its (typically, QN-rich) prion domain is efficient only in the presence of another aggregated (in most cases, QN-rich) protein.
Here, we demonstrate that a fusion of the prion domain of yeast protein Sup35 to some non-QN-rich mammalian proteins, associated with amyloid diseases, promotes nucleation of Sup35 prions in the absence of pre-existing aggregates. In contrast, both a fusion of the Sup35 prion domain to a multimeric nonamyloidogenic protein and the expression of a mammalian amyloidogenic protein that is not fused to the Sup35 prion domain failed to promote prion nucleation, further indicating that physical linkage of a mammalian amyloidogenic protein to the prion domain of a yeast protein is required for the nucleation of a yeast prion. Biochemical and cytological approaches confirmed the nucleation of protein aggregates in the yeast cell. Sequence alterations antagonizing or enhancing amyloidogenicity of human amyloid-␤ (associated with Alzheimer's disease) and mouse prion protein (associated with prion diseases), respectively, antagonized or enhanced nucleation of a yeast prion by these proteins. The yeast-based prion nucleation assay, developed in our work, can be employed for mutational dissection of amyloidogenic proteins. We anticipate that it will aid in the identification of chemicals that influence initial amyloid nucleation and in searching for new amyloidogenic proteins in a variety of proteomes.
Cross-␤ fibrous protein polymers, reproduced and spread via nucleated polymerization and termed amyloids, are associated with a variety of human and animal diseases, including Alzheimer's (AD), 8 Parkinson's (PD), and Huntington's (HD) diseases (1). Some evidence points to the connection between amyloids and type II diabetes (2,3). Infectious amyloidogenic prion protein (PrP) is responsible for transmissible spongiform encephalopathies (TSEs), also called prion diseases (4,5). Some amyloid and prion diseases (including AD and TSEs) are fatal, and most of them are incurable. AD is typically reported as the 6th most frequent cause of death in the United States (6); however, this is certainly an underestimate, as AD was routinely underdiagnosed in the past, and a significant portion of Americans over the age of 65 are dying from complications caused by AD. Therefore, estimates evaluating AD as the 3rd most frequent cause of death in the United States (7), and possibly in other developed countries with a long life-expectancy, are likely to be more realistic. Moreover, health-care costs related to dementias (i.e. mostly to AD) are estimated at the level of $259 billion in the year 2016 (6), and AD is one of the major factors affecting the quality of life at an advanced age (8). In case of TSEs, the epidemics of "mad cow" disease led to huge losses in the European cattle industry (5). In addition to disease-related amyloids and prions, some amyloids or amyloid-like protein polymers have been described that are related to biologically positive functions, such as formation of structural fibrous materials (e.g. silks), storage of peptide hormones, scaffolding of covalent polymers (e.g. melanin), and long-term memory (9,10).
Despite such a broad biological impact of amyloids and prions, the mechanism of their initial formation in vivo remains a mystery. Many proteins possess an amyloid-forming potential in the test tube (1); however, it is effectively suppressed in the biological systems, apparently due to its interference with the normal (functional) protein fold. In most cases, it is unclear how disease-related or functional amyloids overcome such a suppression. Some amyloid and prion diseases are caused by mutations that either occur in the gene coding for amyloidforming protein or influence production of this protein (11)(12)(13), although the molecular basis of the impact of these mutations on protein structure remains poorly understood in most cases. However, a majority of amyloid diseases (including the vast majority of AD cases) occur sporadically, due to unknown reasons (14). Problems with structural studies of amyloidogenic proteins (e.g. due to their inability to form crystals) and differences between patterns of amyloid formation in vitro (where most proteins can undergo this process depending on conditions) and in vivo (where only a few of them are capable of doing so) further complicate the understanding of the initial steps of amyloid nucleation in natural conditions. Low rates of in vivo amyloid formation and late onset or long incubation periods, detected for most amyloid and prion diseases, make it difficult or impossible to catch amyloid nucleation in mammalian models. As a variety of unrelated proteins with very different sequences are shown to form amyloids, the sequence requirements for amyloid nucleation also remain elusive.
In yeast and other fungi, endogenous prions (in most cases of amyloid nature) manifest themselves as non-Mendelian elements heritable via cytoplasm (15,16). Yeast prion proteins contain so-called prion domains (PrDs) that are entirely responsible for the intermolecular interaction leading to the formation of an amyloid axis, and are, at least in some cases, distinct from domains responsible for the major cellular functions of the same proteins. Some yeast prions control easily detectable phenotypic traits, typically resulting from a partial loss of the cellular function of a protein as a result of its incorporation into prion polymers. For example, formation of a prion state (termed [PSI ϩ ]) by the Sup35 protein, a yeast counterpart of a eukaryotic translation termination factor eRF3 (17,18), results in defective termination of translation, leading to a read-through of stop codons (nonsense-suppression), an easily detectable phenotype in specially designed yeast strains (15). This enables researchers to develop phenotypic assays for prion formation that are capable of detecting even rare events.
By using yeast models, it was demonstrated that de novo prion nucleation is induced by the transient overproduction of a prion-forming protein or its PrD (19 -22). However, an efficient prion induction by overproduced Sup35 protein or its PrD requires the presence of another prion, usually [PIN ϩ ] (or [RNQ ϩ ]), a prion form of Rnq1 protein (23)(24)(25). It was proposed that Rnq1 prion polymers nucleate the initial assembly of the Sup35 polymers. A transient direct association between Rnq1 polymers and Sup35 appears likely as the prion domains of both these proteins are rich in Gln and Asn residues. This is also true for the majority of other yeast prion proteins, and the presence of the other (in most cases, QN-rich) protein in an aggregated form was indeed shown to reproduce the effect of Rnq1 prion on [PSI ϩ ] induction (24,25). It is also possible to promote de novo nucleation of the Sup35 prion in the absence of Rnq1 prion by expressing a modified PrD of Sup35. For example, an attachment of either an artificial highly hydrophobic extension (23) or a polyQ region of the same length as in the human huntingtin protein in case of HD (26) to the Sup35 PrDcontaining derivative promoted [PIN ϩ ]-independent prion formation. However, it remained unclear whether known amyloidogenic proteins or domains, which are not QN-rich, can influence prion nucleation by a QN-rich yeast protein.
Here, we show that the initial nucleation of Sup35 protein in the absence of other pre-existing prions can be promoted by the attachment of a mammalian amyloidogenic protein (or domain) to Sup35 PrD. Moreover, we demonstrate that a de novo prion nucleation assay in yeast can be applied to dissecting the role of specific regions or amino acid residues of mammalian proteins in altering the amyloidogenic properties of these proteins

Mammalian amyloidogenic proteins do not promote [PSI ؉ ] nucleation in trans
It has previously been shown that co-overproduction of some yeast prionogenic proteins can promote prion formation by another yeast prion protein in the strain lacking pre-existing prions (24,25,27,28). To determine whether mammalian amyloidogenic proteins exhibit such an effect on prion formation by the yeast protein Sup35, we have overproduced the mouse prion protein (PrP), associated with TSEs, and the human ␤-amyloid peptide (A␤), associated with AD, in a yeast strain lacking pre-existing prions ([pin Ϫ psi Ϫ ]) either individually or simultaneously with separately expressed Sup35 PrD, Sup35N ( Fig. 1). In the case of PrP, we have employed the region between positions 90 and 230, which is sufficient to generate and maintain a prion state in mammals (29). In the case of A␤, the most amyloidogenic and pathogenic (30) 42-residue variant (A␤(1-42)) has been employed. We have confirmed that the PrP(90 -230) protein is produced in yeast (Fig. S1). However, the levels of A␤(1-42) were below detection limits (data not shown), possibly due to a low proteolytic stability of this short peptide in yeast cells. Therefore, we have also used the PrP(90 -231)-GFP and A␤(1-42)-GFP fusion proteins that are both produced at high levels in yeast cells (Fig. S1A). The PrP-and A␤-based constructs produced amyloid-like detergent-resistant aggregates in yeast, as shown previously (31) and confirmed by us (Fig. S1B) using the semi-denaturing detergent agarose gel electrophoresis, SDD-AGE (32). To detect [PSI ϩ ] formation, we employed the ade1-14 (UGA) reporter (15). The [psi Ϫ ] strains bearing this reporter are Ade Ϫ (i.e. they do not grow on medium lacking adenine) and only rarely produce spontaneous Ade ϩ colonies, in part due to reversions or suppressor mutations. The conversion of endogenous Sup35 into a prion form leads to a termination defect and read-through of ade1-14, resulting in an Ade ϩ phenotype. Therefore, [PSI ϩ ] induction

Nucleation of a yeast prion by mammalian proteins
can be detected as an increase in the frequency of Ade ϩ colonies over a low background. None of the mammalian proteins (PrP(90 -230), PrP(90 -231)-GFP, A␤(1-42), or A␤(1-42)-GFP) was able to induce [PSI ϩ ] formation, both at normal levels of Sup35 and in the presence of excess Sup35N (Fig. 1, A and B). This was in contrast to the yeast prionogenic QN-rich protein Lsb2 (fused to GFP) that promoted [PSI ϩ ] formation in the presence of excess Sup35N (Fig. 1, A and B), as shown previously (23,33,34).

Mammalian amyloidogenic proteins promote [PSI ؉ ] nucleation when attached to the Sup35 prion domain
Next, we checked what happens if a mammalian amyloidogenic protein is physically attached to the fragment of Sup35 containing the PrD. We prepared a series of such constructs as shown in Fig. 2A. Some of them also contained an HA tag (see "Experimental procedures"), which does not influence the ability of the protein to induce a prion, according to our data (not shown). The Sup35N fragment (roughly equivalent to Sup35 PrD), produced from a copper-inducible (P CUP1 ) promoter, can slightly induce the formation of Ade ϩ ([PSI ϩ ]) colonies in a [pin Ϫ ] strain at high concentrations of CuSO 4 (e.g. see Table 1), but this effect is weak and is not clearly detectable in plate assays (e.g. see Fig. 2B) unless very high concentrations of CuSO 4 and/or very long incubation periods are used.
No [PSI ϩ ] induction occurs when Sup35N alone is expressed from the galactose-inducible P GAL promoter (e.g. see Fig. 1A). The Sup35NM fragment, bearing both PrD and the middle region (Sup35M), which contains a high concentration of charged residues and is responsible for keeping Sup35 in a soluble state, does not induce the formation of Ade ϩ colonies in the [pin Ϫ ] strain (e.g. see Fig. 2C and Table 1). Notably, an attachment of the region coding for either mouse PrP (90 -230) or human A␤  to the C terminus of Sup35N ( Fig. 2A) enabled such a chimeric construct to induce Ade ϩ colonies after incubation in the presence of CuSO 4 even in conditions

Nucleation of a yeast prion by mammalian proteins
when induction by Sup35N alone was not detectable (see Fig.  2B, Fig. S2, and Table 1). [PSI ϩ ] induction by Sup35N-A␤  was stronger than that by Sup35N-PrP (90 -230), and it could be detected even at background levels of CuSO 4 as seen in the quantitative assay (Table 1) and occasionally in the plate assay (e.g. Fig. 2B). More importantly, the ability of these chimeric constructs to induce Ade ϩ colonies was not promoterspecific and did not depend on the presence of CuSO 4 per se, as it was reproduced by using the chimeric constructs expressed from the P GAL promoter (Fig. 2C). A␤(1-42) also promoted Ade ϩ formation when fused to the Sup35NM fragment and was expressed from either a P GAL (Fig. 2C) or P CUP1 (Fig. 2D) promoter, albeit at a lower frequency (Table 1), and in the latter case at higher concentrations of CuSO 4 when compared with Sup35N-A␤ . However, we have not detected Ade ϩ induction by the Sup35NM-PrP(90 -230) construct (Fig. 2D).
We have shown that the vast majority of Ade ϩ colonies induced by the PrP-or A␤-based chimeric constructs in the [pin Ϫ ] strain retain suppression after the loss of the inducing plasmid and are curable by serial passages on medium containing an antiprion agent guanidine hydrochloride (GdnHCl) ( Table 2). These data confirm that the majority of these colonies arise from the conversion of the endogenous Sup35 protein into a prion form, [PSI ϩ ]. Notably, a high expression of the PrPor A␤-based constructs (fused to either GFP or Sup35 PrD) did not inhibit the growth of the [pin Ϫ ] yeast strain (data not shown) and did not increase the levels of the stress-inducible chaperone Hsp104 (Fig. S3A). These results show that prion induction by the PrP or A␤ constructs is not a consequence of proteotoxic stress. As expected, both PrP-or A␤-based chimeric constructs, as well as control Sup35N protein, efficiently induced [PSI ϩ ] formation in a [PIN ϩ ] strain ( Fig S3, B and C).
The ability to promote [PSI ϩ ] nucleation in the absence of [PIN ϩ ] upon fusion to Sup35N or NM is not restricted only to A␤  or PrP (90 -230), as two other human amyloidogenic peptides, namely the aggregation-prone region ("non-amyloid component" or NAC) of ␣-synuclein, associated with PD (14,35), and amylin (or IAPP) peptide, associated with type II diabetes (2), also exhibited such an effect (Fig. 2E). In contrast, several proteins without known amyloidogenic properties, specifically yeast AIR-carboxylase (Ade2), bacterial ␤-galactosidase (LacZ), and jellyfish green fluorescent protein (GFP) did not induce [PSI ϩ ] formation at any noticeable level when fused to Sup35N (Fig. 2F) or NM (Fig. 2G). Notably, some of these proteins are known to form multimers, and in a separate experiment employing the ade2 mutant strain, we have specifically shown that the Sup35N-Ade2 and Sup35NM-Ade2 constructs produce functional AIR carboxylase in yeast (Fig. S3D). This confirms that these chimeric proteins form multimeric com-plexes in yeast, because the functionality of AIR carboxylase depends on its multimerization (36). Therefore, our data show that the ability of a protein to promote prion nucleation in a fusion to a fragment bearing the PrD of Sup35 depends on the amyloidogenic properties of such a protein, rather than with its ability to form multimeric complexes per se.

Mammalian amyloidogenic proteins promote the formation of biochemically and cytologically detectable aggregates in the yeast cells
By using semi-denaturing detergent-agarose gel electrophoresis (SDD-AGE), we have demonstrated that the chimeric proteins containing PrP(90 -230) or A␤(1-42) produce detergentresistant polymers in the yeast cells lacking pre-existing prions (Fig. 3A), as is typical of yeast prions and amyloids (32), and promote the immobilization of endogenous Sup35 protein into an aggregated fraction (Fig. 3B). Likewise, the chimeric protein, containing the NAC region of ␣-synuclein, also produced the detergent-resistant aggregates in yeast in the absence of preexisting prions (Fig. 3A). Thus, phenotypically detectable [PSI ϩ ] formation coincides with physical aggregation of the inducer protein and immobilization of the inducee protein into aggregates.

Nucleation of a yeast prion by mammalian proteins
A high level expression of a fluorescently-tagged Sup35NM fragment is known to produce cytologically detectable aggregates efficiently in a strain bearing pre-existing prions (e.g. [PIN ϩ ], the prion form of Rnq1) but not in a [pin Ϫ ] strain (37,38). These aggregates include both dot-like puncta and filamentous assemblies, seen as lines, ribbons, or rings (see examples in Fig. 3C), and are defined here as "filaments." It is shown (37,38) that filaments specifically represent intermediate stages of the prion generation pathway, whereas dots may correspond to other types of intermediates, to mature prions, and possibly to non-prion aggregates. We have confirmed that Sup35NM-YFP produces detectable aggregates only very rarely and transiently in the [pin Ϫ ] strain (Fig. 3D), and it never produces filaments. In contrast, co-expression of Sup35N-A␤(1-42) with Sup35NM-YFP in the [pin Ϫ ] strain resulted in an efficient accumulation of fluorescently-tagged aggregated structures (Fig.  3D), detectable as early as after 6 h of induction, and increased in numbers with prolonged incubation. These structures included filaments (Fig. 3E) similar to those detected previously after the expression of fluorescently-tagged Sup35NM in the [PIN ϩ ] cells (37,38). The co-expression of Sup35N with Sup35NM-YFP produced cytologically detectable aggregates at a significantly lower frequency and with slower kinetics, compared with Sup35N-A␤(1-42) (Fig. 3D). These data confirm that fusion to A␤ promotes the nucleation of cytologically detectable aggregated structures by Sup35 PrD.

[PSI ؉ ] induction by chimeric proteins is not due to the induction of [PIN ؉ ]
Both Sup35N-PrP(90 -230) and Sup35N-A␤(1-42) constructs were capable of nucleating the [PSI ϩ ] prion in an rnq1⌬ strain, lacking the Rnq1 protein (Fig. 4A) (Fig. 4C). These data show that [PSI ϩ ] nucleation in the pres-ence of chimeric constructs bearing mammalian amyloidogenic proteins is not due to the induction of [PIN ϩ ] prion or other yeast non-Sup35 prions with a similar effect.

A␤ alterations influence prion nucleation in yeast in the same direction as they do in humans or in vitro
Next, we checked whether alterations in A␤ known to influence prion propagation and disease in mammals have similar effects in yeast. Several variants of A␤ peptide exist in humans, of which A␤(1-42) and A␤(1-40) (lacking the last two amino acid residues) are the most abundant ones (41). Of these two, A␤(1-42) is considered to be the most amyloidogenic and most pathogenic form in humans (30,42). In yeast, Sup35N-A␤(1-40) nucleated [PSI ϩ ] much less efficiently than did Sup35N-A␤(1-42) (Fig. 5A), whereas Sup35NM-A␤(1-40) did not nucleate [PSI ϩ ] at all (Fig. 5B). Notably, removal of the two N-terminal amino acid residues of A␤ within the chimeric construct did not inhibit [PSI ϩ ] nucleation (Fig. 5C), in agreement with structural models placing the N-terminal region of A␤ outside of the amyloid core (43,44).
We have also generated several mutations at the positions of A␤(1-42) known to influence amyloid formation. Previous in vitro experiments and structural data identified positions 19, 20, and 31 as being important for amyloid formation by A␤ (45)(46)(47) and located within intramolecular cross-␤-sheets of A␤(1-40) polymers (48,49). However, according to the most recent structural model of A␤(1-42) polymers (44), only position 31 is located within one of the ␤-strands, whereas positions 19 and 20 are involved in hydrophobic interactions. Indeed, the
Phenotypic characterization of the effects of A␤ alterations was confirmed by aggregation assays in yeast.

PrP alterations influence prion nucleation in yeast in the same direction as they influence the formation of PrP prion in mammals
Next, we checked whether correspondence between known effects of sequence alterations on the amyloid formation in other systems and on prion nucleation in yeast also holds true for PrP. Amino acid substitution P101L in mouse PrP (see Fig.  7A) corresponds to the human mutation P102L, associated with a heritable prion disease, and is shown to cause disease accompanied by a production of the infectious PrP protein in mice (10). In contrast, the substitution Q167R is shown to inhibit prion replication in mice (51). In agreement with these data, substitutions P101L and Q167R in the Sup35N-PrP(90 -230) construct, respectively, increased or decreased [PSI ϩ ] nucleation in the yeast assay (Fig. 7B). The region between amino acid residues 90 and 119 is required for the susceptibility to prion disease in mammals (52). We have shown that a deletion of this region knocks out [PSI ϩ ] nucleation by the chimeric Sup35N-PrP protein in yeast ( Fig. 7C and Table 1) and essentially eliminates the immobilization of full-size Sup35 protein into aggregates in the yeast cells as detected by SDD-AGE (Fig.  7D). Truncation of human PrP after positions 144 or 159, eliminating the C-terminal region, leads to a heritable disease with symptoms similar to prion disease (53)(54)(55), and truncated PrP forms amyloids in vitro (56). We have shown that C-terminal truncations of mouse PrP (at positions 144, 159, or 171) in the Sup35N-PrP chimeras significantly increased both [PSI ϩ ] nucleation ( Fig. 7C and Table 1) and immobilization of Sup35 into amyloid aggregates (Fig. 7D), and truncation at position 159 of PrP also enabled [PSI ϩ ] nucleation in a fusion to Sup35NM (Fig. 7E). These data agree with the notion that

Nucleation of a yeast prion by mammalian proteins
C-terminal PrP truncations trigger the formation of disease via nucleating prion-like aggregates, even though transmissibility of such aggregates has not been proven. Notably, the PrP fragment, including only residues from 90 to 119 did not promote [PSI ϩ ] nucleation when fused to Sup35 (Fig. S5), indicating that although this region is essential for prion formation (see above), it is not sufficient for this process. The presence of the N-terminal region of PrP(23-89) increased [PSI ϩ ] nucleation in yeast, as demonstrated by the ability of the chimeric Sup35NM-PrP(23-230) protein to nucleate [PSI ϩ ] (Fig. 7F), in contrast to the Sup35NM-PrP(90 -230) construct (see Fig. 2D). Although the 23-89 region of PrP is not necessary for prion formation or propagation in mammals, it contains oligopeptide repeats, whose expansions are known to cause a heritable disease with symptoms similar to a prion disease (57,58), Overall, our data show that PrP alterations influence its ability to nucleate prions in yeast in the same direction as they influence (or are suggested to influence) prion diseases in mammals and humans.

Effects of chimeric constructs on prion nucleation are not due to alterations in protein levels
One possible explanation for chimeric constructs, as well as for alterations of PrP or A␤ to influence [PSI ϩ ] nucleation, could be through altering the levels of chimeric proteins. To investigate this possibility, we have compared levels of proteins accumulated in yeast cells at the same concentrations of CuSO 4 . As described previously (59), and confirmed by us (Fig.  S4A), Sup35N is accumulated at low levels in yeast, despite the fact that it has a higher prion-inducing activity in comparison with Sup35NM and Sup35. This is probably due to the high misfolding capability and proteolytic instability of Sup35N. The Sup35N-PrP(90 -230) chimeric protein was produced at higher levels compared with Sup35N (Fig. S4A). However, this could not explain the increased prion-nucleating activity of Sup35N-PrP(90 -230), because the Sup35N-PrP(120 -230) derivative, not capable of prion nucleation, was produced at about the same level as Sup35N-PrP(90 -230) (Fig. S4B). Moreover, the C-terminal truncated derivatives of Sup35N-PrP, which exhibited increased [PSI ϩ ] nucleation, were in fact accumulated at lower levels compared with Sup35N-PrP(90 -230) (Fig. S4B). In theA␤-basedseries,theprion-inducingSup35N-A␤(1-42)construct was accumulated at the same levels as the prion noninducing Sup35N-A␤(1-40) and Sup35N-A␤(1-42) triple F19S,F20S,I31P mutant (Fig. S4, C and D), respectively. The [PSI ϩ ]-inducing Sup35NM-A␤(1-42) construct was accumulated at the same level as the non-inducing Sup35NM-A␤(1-40) construct, and both were less abundant than the non-inducing control, Sup35NM (Fig. S4E). Overall, our data show that although cellular levels of proteins used in this work could vary in some cases, the differences in prion nucleation cannot be explained by differences in protein abundance.

Different chimeric constructs induce different spectra of prion strains
The Sup35 protein can produce a variety of prion variants or "strains" that presumably correspond to various amyloid structures (20, 37). These strains can be differentiated from each other based on both their phenotypic manifestations and biochemical patterns. "Stronger" strains are characterized by higher levels of nonsense codon read-through (leading to better growth on ϪAde medium and lighter color on complete medium in the case of the ade1-14 reporter) and by higher mitotic stability compared with "weaker" strains. This is due to the fact that "stronger" strains are generated by amyloid fibrils with a less rigid amyloid core that are more efficiently fragmented by the yeast chaperone machinery and therefore produce a larger number of oligomeric "seeds," making immobilization of newly synthesized Sup35 and proliferation of prion state more efficient (60). Once established, the prion strain typically faithfully reproduces its observable characteristics.

Nucleation of a yeast prion by mammalian proteins
To determine whether mammalian amyloidogenic proteins influence the parameters of prion strains produced in yeast, we compared spectra of prion strains generated in the presence of different inducing constructs. For this purpose, [PSI ϩ ] isolates were divided into three groups designated as "strong," "intermediate," and "weak" strains on the basis of growth on ϪAde medium and color on complete (YPD) medium. Strain patterns were scored after elimination of the inducing plasmid, in order to exclude the possibility that the continuous presence of a chimeric construct influences the phenotypic manifestation of a [PSI ϩ ] strain. Data are shown in Fig. 8, Fig. S6, and in Table  S1. Although Sup35N-PrP(90 -230) and Sup35N-A␤(1-42) induced preferentially or exclusively "strong" strains, the constructs with some deletion PrP derivatives, such as Sup35N-PrP (90 -159)  Altogether, these results indicate that the preferable type of a yeast prion strain, at least in part, depends on the mammalian amyloidogenic protein used in the inducing construct.

[PIN ؉ ]-independent [PSI ؉ ] nucleation
Efficient prion nucleation by the overproduced yeast Sup35 protein or its PrD-containing fragments typically requires the presence of another (usually QN-rich) protein in an aggregated form (23)(24)(25). A fusion of some Sup35 PrD-containing derivatives to extended polyQ tracts, resembling those associated with HD in humans, or to a yeast prion-forming protein Rnq1 promotes nucleation of the Sup35 prion even in the absence of pre-existing Q/N-rich yeast prions (26,61). However, expanded polyQ constructs and QN-rich proteins were also reported to promote Sup35 aggregation in trans (62), so that an addition of a polyQ or another QN-rich region to the QN-rich Sup35 PrD could be interpreted as an expansion of Sup35 PrD. Our new data demonstrate (to our knowledge, for the first time) that a fusion of the Sup35 PrD-containing region (Sup35N or Sup35NM) to a non-QN-rich mammalian protein (or protein domain) with proven amyloidogenic properties is sufficient for nucleating the formation of Sup35-based prions in yeast cells lacking known pre-existing prions. An apparent explanation for this result is that mammalian proteins nucleate an amyloid in yeast, thus promoting amyloidization of the attached yeast prion domain (Fig. 9). This leads to immobilization of fulllength endogenous yeast protein into prion aggregates, thus allowing for phenotypic detection of a yeast prion. Importantly, a covalent attachment of a mammalian "inducer" to Sup35N (or NM) is required for prion nucleation, as mammalian non-QNrich amyloidogenic proteins do not promote [PSI ϩ ] induction in trans (Fig. 1). As expected, the Sup35N-based chimeric proteins are more efficient in prion nucleation than the Sup35NMbased chimeric proteins, apparently due to an anti-nucleation effect of the M region of Sup35, which contains stretches of potentially repulsive charged residues. This explains why the previous work by Choe et al. (63) failed to detect [PSI ϩ ] induc-  Fig. S6 and Table S1.

Role of protein amyloidogenicity in [PSI ؉ ] nucleation
Importantly, the formation of amyloid nuclei is not triggered by just any kind of protein multimerization, as shown by the lack of [PSI ϩ ] induction in the presence of chimeric constructs, producing non-amyloidogenic multimeric proteins such as such as Ade2 and LacZ (Fig. 2, F and G). Fusions of Sup35N with mammalian amyloidogenic proteins are characterized by higher protein abundance at the same levels of expression, compared with proteolytically unstable Sup35N (Fig. S4A). However, the increased frequency of prion nucleation by Sup35N-based chimeric proteins is not simply due to an increase in the abundance of chimeric constructs. Neither the deletion of the 90 -119-region of PrP in Sup35N-PrP nor the triple substitution at positions 19, 20, and 31 of A␤ in Sup35N-A␤ decreased the abundance of chimeric constructs (Fig. S4, B and D); however, they knocked out the ability of a chimeric protein to nucleate [PSI ϩ ] (Figs. 5D and 7C). Moreover, the Sup35NM construct is expressed at even higher levels than Sup35NM-based prion-inducing chimeric proteins; however, it cannot nucleate [PSI ϩ ] in the [pin Ϫ ] cells. This shows that the increased prion nucleation by chimeric constructs is a result of their amyloidogenic properties, leading to the initiation of the self-assembly into an amyloid form.

Amino acid residues influencing prion nucleation by A␤
In the case of A␤ peptide, data from the yeast assay are also in a good agreement with existing results obtained in other systems. For example, the A␤(1-40) peptide lacking the two C-terminal hydrophobic amino acids, Ile-41 and Ala-42, is considered to be less aggregation-prone and is a typically non-pathogenic A␤ isoform in humans (30,42). This peptide is drastically inefficient in prion nucleation in the yeast assay, compared with the highly amyloidogenic and presumably pathogenic A␤(1-42) (Fig. 5A). Although previous structural studies used the in vitro-produced A␤40 polymers (48,49), the high-resolution structures of A␤(1-42) amyloids, mostly based on solidstate NMR, have also been reported recently (43,44,50). These structures include two molecules per polymer unit, and five intermolecular ␤ sheets spanning residues 2-6 (␤1), 15-18 (␤2), 26 -28 (␤3), 30 -32 (␤4), and 39 -42 (␤5) per each "half" of the fibril. The anti-nucleation effects (Fig. 5, D and E) of substitutions I31P (breaking a ␤4 strand), F19S (disrupting hydrophobic interactions with the ␤2 strand), and K28E (affecting the ␤3 strand and disrupting a "salt bridge" potentially involved in stabilization of an amyloid structure) in the yeast assay are in good agreement with the published structural models. Likewise, the pro-nucleation effect (Fig. 5F) of the D23N substitution, corresponding to a so-called "Iowa mutation," a heritable case of AD (9,11), is also in good agreement with the models. This substitution removes one of the negatively charged residues presumably facing the solvent that might increase an aggregation propensity.

Sequence requirements for prion nucleation by PrP
The region between residues 90 and 119 of PrP, which is known to be essential for the susceptibility to prion infection in mammals (52), is also required for prion nucleation in yeast, although the N-terminal region of PrP(23-89) is dispensable for both (Fig. 7). Mutation P101L, associated with heritable prion disease in mammals (10), increased, whereas mutation Q167R, inhibiting prion replication in mammals, decreased PrP-dependent prion nucleation in the yeast assay. Increased prion-nucleating ability of the fragments lacking the C-proximal region of PrP (Fig. 7) agrees with previous reports linking C-proximal PrP truncations to a heritable neurological disease in humans (53)(54)(55), and it supports a notion that this disease is likely to be prion-like in nature. One possible explanation for this effect is that the ␣-rich C-proximal domain of PrP stabilizes the native conformation and therefore antagonizes the initiation of the cross-␤ (prion) conformation (64,65). Although the structural organization of PrP in a prion form remains a matter of debate (66 -68), our data agree with models locating cross-␤ interactions within the region 90 -170, suggesting the retention of the native secondary structure by the C-terminal region of PrP (68 -70), and predict that the proposed ␤-structure at positions 160 -164 is dispensable for prion initiation. However, our data do not necessarily contradict a possibility of further expansion of the amyloid core to the C-proximal region as shown for some PrP-based amyloids (71,72). Most importantly, our yeast assay provides a tool that could be employed to further decipher sequential and structural requirements for initiation of PrP polymerization and conformational conversion.

Impact of a nucleating construct on spectra of induced [PSI ؉ ] "strains"
Both yeast (13) and mammalian (73)(74)(75) prion and amyloid proteins are known to form various variants or strains that differ from each other by phenotypic and biochemical characteristics and are apparently controlled by distinct protein conformations. Interestingly, we have found out that the spectra of [PSI ϩ ] strains induced by different chimeric constructs differ from each other (Fig. 8). One possible explanation for these data is the formation of distinct initial nuclei by different attached regions of chimeric proteins, followed by an expansion of the amyloid region to different regions of the attached Sup35N domain. Such a mechanism would correspond to a "deformed templating" model previously proposed for strain conversions in PrP prions (76). In this scenario, the spectra of [PSI ϩ ] strains might corroborate to the differences in the "hybrid" templates formed by the fusion proteins. An alternative explanation is that certain strain conformations formed by Sup35N are more compatible, whereas other strain conformations are less compatible with an amyloid conformation formed by a specific mammalian amyloidogenic protein physically attached to the same molecule.

Potential applications of the yeast prion nucleation assay
As the ability of a chimeric construct to nucleate prion formation clearly depends on the amyloidogenicity of the attached protein, our assay can be used to investigate amyloidogenic and prionogenic properties of mammalian and human proteins.

Nucleation of a yeast prion by mammalian proteins
Although several yeast assays for A␤ were proposed previously (31,(77)(78)(79), none of them specifically address the initial nucleation of A␤ polymerization. Our approach could be used to map regions and identify amino acid residues specifically important for polymer nucleation, a crucial step triggering the subsequent amyloid formation and pathogenicity of A␤, PrP, and other disease-related amyloidogenic proteins. Moreover, this assay can be employed to search for chemical factors and conditions specifically modulating the process of initial amyloid nucleation in both a general and a protein-specific manner. This may pave the way for the development of both therapeutic and prophylactic treatments for amyloid diseases that address a triggering mechanism of the disease, initial amyloid nucleation. The major advantage of our system in comparison with previously proposed yeast-based and cell-based assays is that our assay does not require the chimeric fusion protein to propagate a prion state in yeast. Prion detection is achieved by transferring the amyloid state to the endogenous yeast Sup35 protein, so that even transient amyloid formation by a chimeric construct is then fixed and amplified by conversion of an endogenous yeast protein into a prion. Furthermore, non-amyloid multimeric proteins are apparently not capable of nucleating prion formation at high efficiency in our system, making it possible to use this assay for identifying new potentially amyloidogenic proteins or domains, originating from various organisms, including humans. The rapid and easy phenotypic detection of prion nucleation in yeast makes our assay amenable to highthroughput approaches. It is, of course, possible that some sequences identified in such a way would be amyloidogenic only within the context of a chimeric construct containing the prion domain of Sup35, rather than on their own, and/or that some non-canonical assemblies (other than typical amyloids) may facilitate amyloid nucleation. For example, a short artificial extension of highly hydrophobic contents that makes the Sup35N-containing fragment capable of nucleating [PSI ϩ ] in the absence of other known prions (23) may fall into this category. However, even such possible "false positives" might in fact turn out to be useful for a better understanding of the mechanism of initial amyloid/prion nucleation in the cell.

Yeast strains
The Saccharomyces cerevisiae strains used in this study are listed in Table S2. Haploid [PSI ϩ PIN ϩ ] strains GT81-1C and GT81-1D are meiotic spore clones of the homozygous (except mating type) autodiploid GT81 (80). The [psi Ϫ pin Ϫ ] strains GT409 and GT197 were obtained, respectively, from GT81-1C and GT81-1D via curing them of [PSI ϩ ] by guanidinium hydrochloride (GdnHCl), whereas the [psi Ϫ PIN ϩ ] strain GT159 was obtained via curing GT81-1C of [PSI ϩ ] using excess Hsp104. The rnq1⌬ strain GT564 was obtained by K. Gokhale in the Chernoff laboratory via replacing the RNQ1 gene with the Schizosaccharomyces pombe ortholog of the HIS3 gene in the strain GT159. To make sure that our results are not strainspecific, we have also checked [PSI ϩ ] induction by some chimeric constructs in the [psi Ϫ pin Ϫ ] strain GT17 of the 74-D694 genotype (19,20,23). Results obtained with this strain were similar to those seen with GT409. Strain 33G-D373, as described previously (81) and containing a double point mutation in the ADE2 gene, was used for determining the functionality of the Ade2-based chimeric proteins. Prototype "strong" ( ϩ 1-74-D694 or OT56) and "weak" ( ϩ 7-74-D694 or OT56) strains, obtained as described earlier (20), and used in our previous work (82), were employed for the phenotypic comparisons with [PSI ϩ ] strains, induced by chimeric constructs.

Plasmids and primers
The S. cerevisiae-Escherichia coli shuttle plasmids used in this study and primers used in plasmid constructions are shown in Tables S3 and S4 respectively. The DNA regions coding for Sup35NM (with HA tag) and PrP (90 -230) were initially inserted in the pcDNA3.1/Zeo(ϩ) backbone; the chimeric genes coding for Sup35N-PrP(90 -230), Sup35NM-PrP (90 -230), and Sup35NM-PrP(120 -230) were initially generated in pcDNA3.1/Zeo(ϩ) as well (83). Then, respective constructs were excised by using restriction endonucleases BamHI and XbaI or SacI and inserted under the copper-inducible promoter (P CUP1 ) into a respective centromeric shuttle vector with the URA3 marker. The plasmid pmCUP1-Sup35N-PrP(120 -230) was constructed via replacing the EcoRI fragment that contains the P CUP1 -SUP35NM fragment from the plasmid pmCUP1-Sup35NM-PrP(120 -230) with the EcoRI fragment that contains P CUP1 -SUP35N fragment from the plasmid pmCUP1-Sup35N-PrP (90 -230). The pmCUP1-Sup35N plasmid was constructed by inserting the PCR-amplified BamHI-SacI fragment that contains the SUP35N region from the plasmid Sup35N-PrP(90 -230) into the pmCUP1 vector at the position following the P CUP1 promoter. The genes coding for Sup35N-PrP(90 -119), Sup35N-PrP(90 -144), Sup35N-PrP (90 -159), and Sup35N-PrP (90 -171) were constructed by inserting the PCR-amplified BamHI-XbaI fragments that code for the respective PrP domains, from the plasmid pmCUP1-Sup35N-PrP(90 -230) into the pmCUP1 vector at the position following the P CUP1 promoter. Constructs coding for the HA-tagged derivatives of the Sup35N and Sup35N-PrP(90 -230) proteins were produced by PCR-amplifying the BamHI-SacI fragments, coding for respective proteins, from pmCUP1-Sup35N-PrP(90 -230) with primers adding an HA tag coding sequence to a C-terminal end of each fragment, and inserting the resulting constructs into the pmCUP1 vector at the position following the P CUP1 promoter. Both HA-tagged and non-tagged constructs produced the same results in the [PSI ϩ ] induction assays. The chimeric gene coding for Sup35NM-PrP (90 -159) was constructed by replacing the EcoRI fragment that contains the P CUP1 -SUP35N cassette from the plasmid pmCUP1-Sup35N-PrP (90 -159), with the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35NM-PrP (90 -230). The chimeric gene coding for Sup35N-PrP  was constructed by inserting the PCRamplified BamHI-XbaI fragment that codes for the region 23-230 of PrP from plasmid mPrPcyto (83) into the pmCUP1-Sup35NM-PrP (90 -230) vector at the position following the Sup35NM-coding sequence, replacing the PrP(90 -230)-coding fragment. The gene coding for Sup35NM-A␤(1-42) was constructed by inserting the PCR-amplified EcoRI-NotI Nucleation of a yeast prion by mammalian proteins fragment that contains A␤(1-42) from the plasmid pcDNA3.1(ϩ)-A␤42 (kindly provided by Dr. K. Ugen, University of South Florida) containing the human A␤(1-42)-coding sequence (84) into the pmCUP1-Sup35NM-PrP (90 -230) vector at the position following the Sup35NM-coding sequence, replacing the PrP(90 -230)-coding fragment. The DNA sequence coding for A␤  was placed under the P CUP1 promoter by inserting the PCR-amplified BamHI-XbaI fragment that codes for A␤ , from the plasmid pmCUP1-Sup35NM-A␤ , into the pmCUP1 vector at the position following the P CUP1 promoter. The genes coding for Sup35N-A␤  and Sup35N-A␤  were constructed by replacing the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35NM-A␤42, with the EcoRI fragment that contains P CUP1 -SUP35N from the plasmid pmCUP1-Sup35N-PrP (90 -230). The digestion of an additional EcoRI site at the 3rd codon of A␤  resulted in the generation of pmCUP1-Sup35N-A␤ , although pmCUP1-Sup35N-A␤(1-42) was generated by incomplete digestion. To construct the series of plasmids that are more convenient for construction procedures using the EcoRI digestion, the pmCUP1 vector was digested with EcoRI; the resulting 5Ј-overhang was blunted using mung bean nuclease and religated with the same vector to disrupt the EcoRI site upstream from the sequence coding for P CUP1 . This plasmid, named pmCUP1-nERI, was used to construct pmCUP1-nERI-Sup35N-A␤  by inserting the PCR-amplified BamHI-XbaI fragment that contains the Sup35N-A␤(3-42)-coding sequence from the plasmid pmCUP1-Sup35N-A␤ , into the pmCUP1-nERI vector at the position following the P CUP1 promoter. To disrupt an additional EcoRI recognition site spanning nucleotide positions 7-12 of A␤(1-42)-coding sequence without changing the amino acid sequence, the 3rd codon of A␤(1-42) (GAA) that codes for glutamic acid was mutated to the synonymic codon GAG, and the PCR-amplified EcoRI-XbaI fragment containing the A␤(1-42)-coding sequence with respective change (A␤m(1-42)) was inserted into the plasmid pmCUP1-nERI-Sup35N at the position following the sequence coding for Sup35N. In the [PSI ϩ ] induction assays, the Sup35N-A␤m(1-42) construct produced results similar to the unmodified Sup35N-A␤(1-42) construct. The gene coding for Sup35N-A␤  was constructed by inserting the PCR-amplified BamHI-XbaI fragment, which contains A␤(1-40)-coding sequence from the plasmid pmCUP1-nERI-Sup35N-A␤m , into the plasmid pmCUP1-nERI-Sup35N-A␤m  at the position following the sequence coding for Sup35N. The gene coding for Sup35NM-A␤  was constructed by replacing the EcoRI fragment, which contains P CUP1 -SUP35N cassette, in the plasmid pmCUP1-Sup35N-A␤  with the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35NM-PrP (90 -230). Individual base substitutions in the pmCUP1-Sup35N-PrP(90-230) and pmCUP1-nERI-Sup35N-A␤m(1-42) plasmids were generated in the A␤(1-42)-coding sequence using the QuikChange site-directed mutagenesis protocol (Agilent Technologies, Santa Clara, CA). A construct coding for A␤  with triple amino acid substitution F19S,F20S,I31P was generated by O. Laur at Emory Custom Cloning Core Facility in the plasmid pmCUP1-Sup35NM-A␤ . The genes coding for Sup35N-NAC or Sup35N-IAPP were constructed by inserting the PCR-amplified EcoRI-NotI fragment that contains NAC-HA and IAPP regions from the plasmid p106.NAC, containing the human NAC-coding sequence that corresponds to codons 61-95 of ␣-synuclein gene, and from the plasmid pJ201: 66979-IAPP2_optSc, containing the codons 41-70 of human IAPP-coding sequence that corresponds to residues 8 -37 in mature amylin, respectively, into the plasmid pmCUP1-nERI-Sup35N-A␤  at the position following the sequence coding for Sup35N, replacing the A␤(1-42)-coding fragment. Original plasmids p106.NAC and pJ201:66979-IAPP2_optSc were kindly provided by Dr. V. Conticello from Emory University. The plasmids pmCUP1-Sup35NM-NAC and pmCUP1-Sup35NM-IAPP were constructed by inserting the PCR-amplified EcoRI-NotI fragments that contain NAC-HA and IAPP regions from pmCUP1-Sup35N-NAC and pmCUP1-Sup35N-IAPP and inserted into the plasmid pcDNA3.1(Zeo)-Sup35NM-PrP(90 -230) at the position following the Sup35NMcoding sequence, replacing the PrP(90 -230)-coding fragment. Then, respective chimeric genes were cut from plasmids pcDNA3.1(Zeo)-Sup35NM-NAC and pcDNA3.1(Zeo)-Sup35NM-IAPP with BamHI and XbaI and inserted into the pmCUP1 vector at the position following the P CUP1 promoter. The plasmid coding for the C-terminal fusion of Lsb2 with GFP, expressed under the P CUP1 promoter in the pRS316 backbone, was constructed earlier (31). The chimeric gene coding for Sup35N-LacZ was constructed by inserting the PCR-amplified EcoRI-XbaI fragment that contains the lacZ-coding sequence from the plasmid pSVA1 (kindly provided by Dr. M. D. Ter-Avanesyan, Moscow, Russia) into the plasmid pmCUP1-nERI-Sup35N-A␤m  at the position following the Sup35Ncoding sequence, replacing the A␤(1-42)-coding fragment. The gene coding for Sup35NM-LacZ was constructed by replacing the EcoRI fragment of the plasmid pmCUP1-Sup35N-LacZ that contains the PCUP1-SUP35N cassette with the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35NM-PrP(90 -230). The gene coding for Sup35N-GFP was constructed by inserting the PCRamplified EcoRI-SacII fragment that contains GFP-coding sequence from the plasmid pmCUP1-NM-GFP (85) into pmCUP1-nERI-Sup35N-A␤m  at the position following the Sup35N-coding sequence, replacing the A␤(1-42)-coding sequence. The gene coding for Sup35NM-Ade2 was constructed by inserting the Ade2-coding fragment from the plasmid pRS316GAL-Sup35NM-Ade2 into the plasmid pmCUP1-Sup35NM-A␤  at the position following the sequence coding for Sup35NM, replacing the A␤(1-42)-coding sequence. The gene coding for Sup35N-Ade2 was constructed by replacing the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35NM-Ade2, with the EcoRI fragment that contains the P CUP1 -SUP35NM cassette from the plasmid pmCUP1-Sup35N-PrP (90 -230). Plasmids with constructs under the P GAL promoter were constructed by inserting the BamHI-XbaI fragments with respective chimeric genes from constructs with P CUP1 promoter into the centromeric HIS3 vector pLA1 (82) under the galactose-inducible promoter P GAL . Plasmid

Nucleation of a yeast prion by mammalian proteins
pLH105, containing the HSP104 gene under the strong constitutive P GPD promoter, was a gift from S. Lindquist quoted earlier (86). Plasmids pLA1-Sup35N and pLA1-Sup35, containing, respectively, P GAL -SUP35N and P GAL -SUP35N expression cassettes, were described earlier (86,87). The plasmid pmCUP1-PrP-GFP(URA3), kindly provided by A. P. Galkin and coding for the PrP(90 -231)-GFP chimeric protein, was described earlier (88).TheplasmidpmCUP1-A␤(1-42)-GFP(URA3)wasconstructed via inserting the DNA fragment, encoding A␤  and obtained from the human brain mRNA by RT-PCR, with the addition of BamHI and SacII sites into the plasmid pmCUP1-GFP (85), digested with BamHI and SacII. Plasmid pNM-YFP, containing the SUP35NM-YFP construct under the control P CUP1 promoter and kindly provided by I. L. Derkatch, is based on the LEU2 vector pRS315 and was described previously (62). All regions that underwent PCR amplification as well as immediate flanking regions were verified by sequencing, performed at Eurofins MWG Operon (Huntsville, AL). Isolation of plasmid DNA from bacteria was performed according to standard procedures (89).

Enzymes and antibodies
Enzymes used for molecular cloning, PCR, ligation, and site-directed mutagenesis, including restriction endonucleases BamHI, EcoRI, XbaI, NotI, SacI, ClaI, XhoI, mung bean nuclease, Taq and Pfu DNA polymerases, and T4 DNA ligase, were purchased from New England Biolabs. The antibodies to Sup35N and Hsp104 (from S. Lindquist laboratory), Sup35M (4A5), and PrP (4H11) have been described previously (82, 90 -92). Antibody to HA was purchased from Covance. Antibody to GFP, Ab-13970, was purchased from Abcam. Antibody to A␤ (6E10, Covance, catalogue number SIG 39320) was a gift of L. Walker (Emory University School of Medicine). Secondary anti-rabbit and anti-rat HRP antibodies were purchased from Sigma. Secondary anti-mouse HRP antibody was purchased from GE Healthcare.

Yeast media and growth conditions
Yeast cultures were grown at 30°C. Standard yeast media and standard procedures for yeast cultivation and phenotypic and biochemical analyses were used (93). Cell counts were performed using a hemocytometer (Brightline). Optical densities of yeast cultures were measured at 600 nm using Shimadzu UV-2450 spectrophotometer. Standard synthetic medium contains 3 M copper sulfate (CuSO 4 ); it was supplemented with 10, 50, 100, or 150 M CuSO 4 as indicated to induce higher expression of P CUP1 promoter. Synthetic media lacking adenine, leucine, or uracil are designated as ϪAde, ϪLeu, and ϪUra, respectively. In all cases when the carbon source is not specifically indicated, 2% glucose (Glu) was used. The synthetic medium containing 2% galactose (Gal) or 2% galactose and 2% raffinose (Gal ϩ Raf) instead of glucose was used to induce the GAL promoter. Organic complete YPD medium containing yeast extract (1%), peptone (2%), and glucose (2%) was used for color detection. Organic YPG medium containing glycerol (3%) instead of glucose was used to identify respiratory-incompetent (Pet Ϫ ) transformants that arose due to loss of mitochondrial DNA during transformation and were eliminated from further analysis. Detection assay for [PSI ϩ ], based on the read-through of the ade1-14 (UGA) allele, that results in growth on ϪAde medium and lighter color on YPD medium is described previously (13) and under the "Results." Liquid cultures were grown with at least a 1:5 liquid/flask volumetric ratio in a shaking incubator (200 -250 rpm). Yeast transformations were performed according to the standard Li ϩ protocol (92). Curing of [PSI ϩ ] by GdnHCl was performed by incubating cultures for three consecutive passages (ϳ20 -40 generations) on YPD plates with 5 mM GdnHCl, followed by streaking out on YPD and checking individual colonies by both color and growth on ϪAde medium. The mating assay for the presence of [PIN ϩ ] or another prion with a [PSI ϩ ]-inducing capability is described in Fig. 4A.

Prion nucleation assays
To check for [PSI ϩ ] nucleation, plasmids bearing chimeric and control genes under the P CUP1 or P GAL promoter were transformed into the yeast [psi Ϫ ] strain. For plate assays, transformants were grown on the media selective for the plasmid (e.g. ϪUra) containing 2% glucose as a carbon source and a background concentration (3 M) of Cu 2ϩ , and then velveteen replica-plated onto the same medium with addition of 0, 10, 50, 100, or 150 M CuSO 4 as specified in the figure legends (for P CUP1 constructs) or onto the same medium with 2% galactose instead of glucose (for P GAL constructs) to induce expression of the chimeric genes. After induction (usually for 2 days), plates were replica-plated to ϪAde medium with glucose and without additional CuSO 4 , where overexpression was turned off. [PSI ϩ ] formation was scored by growth on ϪAde medium, typically after about 10 days of incubation. At least eight (and usually more) independent transformants were checked per each strain/plasmid combination to ensure reproducibility. Transformants carrying the control and experimental plasmids were always compared on one and the same plate. One or two representative transformants for each strain/plasmid combination are shown on the figures. In all cases, there were no differences in growth detected on the completed medium or medium selected for the plasmid (for simplicity, respective images are not shown on most figures).
For semi-quantitative and quantitative measurements of [PSI ϩ ] formation, a pre-culture obtained from a fresh transformant colony was grown in the liquid synthetic medium selective for the plasmid up to OD 600 ϭ 2.5 and then inoculated into the liquid plasmid-selective media with additional CuSO 4 (usually 100 M) at a starting concentration of 10 6 cells/ml. Cultures were incubated at 30°C with shaking, with aliquots taken at desired time points, washed with water, diluted appropriately, and either spotted (as serial decimal dilutions) or plated onto both plasmid-selective medium containing adenine (to count numbers of viable plasmid-containing cells) and plasmid-selective medium lacking adenine (e.g. ϪUraϪAde) to detect [PSI ϩ ]. Frequency of [PSI ϩ ] induction was calculated as a ratio of the number of Ade ϩ colonies to the total number of viable plasmidcontaining cells plated. To ensure accuracy, only dilutions that produced plates with fewer than 500 colonies were counted. For each construct, quantitative assay was repeated with at least three cultures, each originated from an independent transfor-Nucleation of a yeast prion by mammalian proteins mant to ensure reproducibility. Standard deviations were calculated according to Ref. 94. Cultures with prion-inducing and control plasmids were always run in parallel in the same experiment.

Protein isolation and characterization: SDS-PAGE, SDD-AGE, and Western blotting
For isolation of the total yeast protein, cells grown in the liquid medium were collected by centrifugation at 2000 ϫ g for 5 min at 4°C, washed with 300 l of ice-cold lysis buffer (25 mM Tris, pH 7.5, 0.1 M NaCl, 10 mM EDTA, 100 g/ml cycloheximide, 2 mM benzamidine, 20 g/ml leupeptin, 4 g/ml pepstatin A, 1 mM N-ethylmaleimide, 1ϫ protease inhibitor mixture from Roche Diagnostics GmbH, 2 mM phenylmethylsulfonyl fluoride), resuspended in 2 volumes of ice-cold lysis buffer, and mixed to ϳ300 l of acid-washed glass beads. Cells were lysed by vortexing 6 times for 30 s, with at least 1 min on ice in-between each time. Cell debris was removed by centrifugation at 2000 ϫ g for 5 min. The amount of protein in the samples was determined by Bradford reagent (Bio-Rad) and normalized using lysis buffer. The protein samples were boiled for 10 min prior to loading onto SDS polyacrylamide gel. After electrophoresis, proteins were transferred onto Immobilon-P 0.45-m polyvinylidene difluoride blotting membrane (EMD Millipore) or Amersham Biosciences Protran Premium 0.45-m nitrocellulose blotting membrane (GE Healthcare) and reacted to appropriate antibodies. Reaction was detected by using the chemiluminescent detection reagents as described in the GE Healthcare protocols. SDD-AGE followed by transfer to the nitrocellulose membrane was performed according to Ref. 32, with modification (addition of 0.1% SDS to the transfer buffer). Protein concentrations were normalized by the Bradford assay.

Fluorescence microscopy
For the microscopic detection of aggregates, the [psi Ϫ pin Ϫ ] yeast strain was co-transformed with the plasmid producing Sup35NM-YFP and either empty control plasmid pRS316, plasmid producing Sup35N-HA, or plasmid producing Sup35N-A␤ . All the constructs were under the control of the P CUP1 promoter. Cultures were pre-grown overnight in a liquid synthetic medium selective for both plasmids, inoculated at OD 600 ϭ 0.5 into fresh medium of the same composition with the addition of CuSO 4 up to 50 M, and grown with shaking. Cells were examined after specified periods of incubation by using a Leica DM6000B microscope and Leica QWin Standard 3.2.0 software (Leica Microsystems GmbH), at ϫ1000 magnification (ϫ100 objective plus ϫ10 eye, with NA 1.5 for the objective). Images were captured with a Leica DC 500 color digital camera.