The [PSI +] Prion Exists as a Dynamic Cloud of Variants

[PSI +] is an amyloid-based prion of Sup35p, a subunit of the translation termination factor. Prion “strains” or “variants” are amyloids with different conformations of a single protein sequence, conferring different phenotypes, but each relatively faithfully propagated. Wild Saccharomyces cerevisiae isolates have SUP35 alleles that fall into three groups, called reference, Δ19, and E9, with limited transmissibility of [PSI +] between cells expressing these different polymorphs. Here we show that prion transmission pattern between different Sup35 polymorphs is prion variant-dependent. Passage of one prion variant from one Sup35 polymorph to another need not change the prion variant. Surprisingly, simple mitotic growth of a [PSI +] strain results in a spectrum of variant transmission properties among the progeny clones. Even cells that have grown for >150 generations continue to vary in transmission properties, suggesting that simple variant segregation is insufficient to explain the results. Rather, there appears to be continuous generation of a cloud of prion variants, with one or another becoming stochastically dominant, only to be succeeded by a different mixture. We find that among the rare wild isolates containing [PSI +], all indistinguishably “weak” [PSI +], are several different variants based on their transmission efficiencies to other Sup35 alleles. Most show some limitation of transmission, indicating that the evolved wild Sup35 alleles are effective in limiting the spread of [PSI +]. Notably, a “strong [PSI +]” can have any of several different transmission efficiency patterns, showing that “strong” versus “weak” is insufficient to indicate prion variant uniformity.


Introduction
Prions in yeast are a new form of gene, composed of proteins instead of nucleic acids [1]. As such, their inheritance, mutation and segregation are not expected to follow the same rules as the majority DNA/RNA genes. The [PSI + ] prion was first recognized as a non-chromosomal genetic element enhancing the read-thru of the premature termination codon in ade2-1 [2]. Its unusual genetic properties led to its identification as a prion of Sup35p [1], a subunit of the translation termination factor [3,4], specifically an amyloid form (b-sheet-rich filamentous polymer of protein subunits) of the normally soluble Sup35p [5][6][7][8][9]. In the amyloid form, the protein is largely inactive, resulting in increased readthrough of termination codons. Yeast prions are important models for mammalian prion diseases, and for amyloid diseases in general.
Prions can often be transmitted between species, as was first recognized by infectivity of sheep scrapie brain extracts for goats [23]. However, cross-species transmission is inefficient (or com-pletely blocked) as a result of sequence differences between the donor and recipient prion proteins [24]. This phenomenon is called the species barrier, and has also been observed in yeast prions [19,[25][26][27][28][29][30][31]. Wild isolates of S. cerevisiae also show considerable sequence variation in Sup35p sequence [20,32], and these sequence differences produce barriers to transmission of [PSI + ] [20], presumably evolved to protect cells from the detrimental, even lethal, effects of this prion [33,34].
A single prion protein can propagate any of a number of prion variants (called 'prion strains' in mammals), with biological differences due to different self-propagating conformations of the amyloid [9,35,36]. Although there is evidence for conformational differences between prion variants, the nature of those differences is not yet known. In yeast, prion variants differ in intensity of the prion phenotype, stability of prion propagation, interactions with other prions, response of the prion to overproduction or deficiency of various chaperones, and ability to cross species barriers [30,31,[37][38][39][40][41]. Different variants arise during prion generation as a result of some stochastic events occurring in the initial formation of the prion amyloid. Generally, prion variant properties are rather stable, even during propagation in a species different from that in which the prion arose (e.g. [42]).
In a previous report, we demonstrated transmission barriers between Sup35 alleles from wild strains of S. cerevisiae, an 'intraspecies barrier'. These intraspecies barriers are of particular interest since they must operate in nature, when S. cerevisiae strains mate among themselves. Interspecies matings are less efficient than intraspecies matings (e.g., [43]), and diploids formed produce almost no viable meiotic spores [44,45]. In most cases, the intraspecies barriers were incomplete, with occasional transmission between strains with different Sup35 sequences. Were the prions transmitted the same variant as the original, or were they prion 'mutants', heritably changed in their properties? Under selective conditions, prion variant properties may change, a phenomenon first demonstrated in mice [46] and also known in yeast [30,47]. Selection in the presence of a different prion protein sequence, or a drug interacting with amyloid could induce a new prion by inaccurate cross-seeding, and reflect generation of a new prion, rather than propagation of one of several sub-variants already present. Here, we examined variation in prion properties under non-selective conditions, finding evidence for the existence of a 'cloud' of variants with stochastic fluctuation.

Prion variant-specificity of intraspecies transmission barriers
Wild SUP35 alleles fall into three groups: the 'reference' sequence is essentially that of laboratory strains; D19 has a 19 residue (66-84) deletion in the prion domain; E9 is representative of a group with N109S and several polymorphisms in the M domain [20]. Three independent prion variants of the E9 Sup35p (E9A, E9F, E9G) were selected in strain 4828 (Table S1). We tested the transmission of these variants by cytoduction to strain 4830 expressing E9 itself, D19 or reference Sup35. None of these variants were transmitted well into the strain containing the D19 Sup35 polymorph. However, two variants (A, G) propagated very poorly with reference Sup35 sequence, while the other variant (F) was able to efficiently transmit the prion to the reference sequence (Table 1, p,10 210 ). This indicates that intraspecies transmission barriers are variant-specific.
Two of the few E9GRD19 cytoductants that were [PSI + ] ( Table 1) were tested for transmission to strains with different Sup35 polymorphs (Table 2). Each had lost the transmission specificity and were now able to transmit the prion more efficiently into all sequences (  (Table 3). This variant originated in the reference sequence, but when transferred to Sup35D19, is then transferred well to either the reference or the D19 Sup35s, but very poorly to E9 (Table 3). In contrast, either of two E9-originating prions in a D19 host ([PSI + E9G]D19), transfer well to all polymorphs (   (Table 3), but when passed back to the reference sequence, only transmitted 20% to D19 (Table 4). Similarly, the prion passed through E9, and able to transmit to another E9 host at 92% (Table 3), once passed back to the reference host could only transmit 46% to E9 (Table 4).
These results indicate that the predominant variant has changed. But is this change due to mistemplating as the prion passes from Sup35 molecules with one sequence to those with a different sequence, or is there an ensemble of variants present within the population that can be selected based on the specific selection pressure, to be visible with a specific transmission phenotype?

Dynamic cloud of prion variants
If the population contains an array of prion variants from which one or another can be selected, one might expect these to segregate during mitotic growth, much as differently marked plasmids sharing the same replicon or mitochondrial genomes will segregate mitotically, even without exposure to a selective condition. In contrast, if the changes in prion variant are due solely to mistemplating when a prion crosses a transmission barrier to a different sequence, then the transmission pattern should not change substantially even after extensive propagation in the original strain. We designed this experiment to separate the mitotic segregation phase, in which there was no change of Sup35p polymorph, from the transmission phase, in which the test of prion variant is then made by cytoduction to the three Sup35 polymorphs.
We subcloned single colonies of the 779-6A [PSI + ref] yeast strain (reference Sup35p) without selection on K YPD plates for at least 75 generations. Table 5 illustrates our surprising result, that many subclones had transmission profiles considerably different from the parent strain 779-6A. This indicates that there is an ensemble of variants or a prion cloud that has different transmission profiles. We have classified these variants as being type A if they transmit well into reference sequence but poorly into D19 and E9 sequences. Type B transmits well into reference and E9 sequences, but poorly into the D19 sequence. Type C transmits well into D19 and reference sequence, but poorly into the E9 sequence and type D transmits well into all sequences. From this subcloning we now had yeast strains that were carrying prion variants of type B (Y1), type C (Y2) and of type D (Y5). These strains repeatedly display these propagation patterns even after many months in frozen stocks. We then wanted to determine if we had now isolated single variants within the original ensemble so each of three clones, of transmission types B, C and D, were subcloned an additional 75 generations on K YPD plates with ten clones of each tested as before. To our surprise these extensively grown subclones of each of the three types still produced clones with an ensemble of prion variants (Table 6). Even the Y1 strain, which did not initially propagate into the D19 sequence, produced subclones with a variety of transmission profiles.
To determine if the appearance of different predominant variants was due to some unrecognized selective pressure on these strains while propagating on K YPD plates, the subcloning was performed in liquid YPD media maintaining the culture in exponential growth phase throughout. Once cell density reached 0.3 absorbance units at 600 nm the cultures were diluted, transferring only 1000 cells to a fresh culture, a process continued for at least 84 generations. Even under exponential growth phase (Table S2), an array of transmission profiles was observed similar to that in Table 6.
The presence of changed transmission patterns in a majority of the clones without any selection having been applied made it clear that the changes were not due to a chromosomal mutation. Nonetheless, we tested for such a chromosomal change by curing [PSI + ] from Y5 by growth on guanidine, and cytoducing cytoplasm from Y1, Y2 or Y5 into strain 4830 and then 8 cytoductants from each were cytoduced into a rhou derivative of the cured Y5 (Table S3). These cytoductants were then cytoduced into recipients each carrying one of the three SUP35 polymorphs. In each case the transmission pattern followed that of the original Y1, Y2, or Y5 donor of cytoplasm, rather than the Y5 pattern of the recipient (Table S3), confirming that the change was due to a new variant of [PSI + ] and not a chromosomal change. The frequency with which the transmission pattern changed without selection or protein over expression is orders of magnitude higher than for the generation of any new prion, and the fact that the change is one of changing the specificity of transmission to different Sup35p polymorphs proves that it is indeed a change of [PSI + ], and not the generation of some other prion.
To further test the presence of an ensemble of prion variants, one subclone of Y1, which had the same profile as the parent, not being able to transmit into the D19 sequence, was subcloned for an additional 75 generations. As shown in Table S4, subclones were obtained with various profiles some with very good transmission into the D19 sequence containing strain. These results indicate that a single variant had not been selected and that an ensemble or cloud of prion variants must exist with a dynamic propagation pattern under non-selective conditions. Each isolate has a specific transmission pattern, even after frozen storage for many months (Table S5). We infer that during growth, events must allow for a stochastic shift of the ensemble to allow for isolation of variants with specific reproducible transmission patterns.

Wild [PSI + ] transmission
[PSI + ] is rare in wild strains [33], but was found in 9 of 690 wild isolates [49], each expressing the reference Sup35 (ref. [49] and Amy Kelly, personal communication). How do these wild [PSI + ] variants respond to the intraspecies barriers we previously reported [20]? We used both reference sequence and E9 sequence Sup35 fused to GFP and could see dots in the reported wild [PSI + ] strains 5672, UCD#885, UCD#978 and UCD#2534, though infrequently, but not in strains UCD#521, 587, 779, 824, 939 ( Figure S1). To test these strains genetically for nonsense suppression, we crossed the wild strains with strain 4972 (Table  S1), carrying the [PSI + ]-suppressible ade1-14 marker, and tested dissected tetrads to determine if ade1-14 is suppressed. We found that for seven of the wild strains, ade1-14 was weakly suppressed in the segregants, and this suppression could be cured by growth in the presence of guanidine, which is known to cure the [PSI + ] prion. We could not obtain tetrads from diploids formed with strain UCD# 978 and strain 5672 gave poor spore germination.
The transmission of the wild [PSI + ] isolates into cells expressing the Sup35 polymorphs in strains 4828 and 4830 by cytoduction is shown in Table 7. The wild [PSI + ] strains transmit well into the reference sequence, but most showed poor transmission to one or both of the D19 or E9 sequences ( Table 7). All four transmission patterns were observed (Table 7), but all of the isolates were 'weak' [PSI + ] ( Figure 1B). Thus, each of the strains tested transmitted [PSI + ] even though several did not show dots with Sup35NM-GFP. Of course, their presumed independent origin means that these wild isolates are not derived from one prion cloud.

Strong [PSI + ] includes several prion variants
Variants of [PSI + ] may be weak or strong in phenotype, stable or unstable in propagation, and have various responses to deficiency or over expression of chaperones or other cellular components, have different patterns of ability to cross species barriers, and, as shown here, to cross intraspecies transmission barriers. To what extent these various parameters are correlated is largely unknown. We tested the several prion variants derived from the [PSI + ] in strain 779-6A with different transmission patterns for their 'strong' vs 'weak' character ( Figure 1A). We note that, with identical chromosomal genotype, they are indistinguishable in the 'strength' parameter in spite of having substantially different transmission properties. As noted above, the wild [PSI + ] variants are indistinguishably 'weak', but have different transmission patterns to the Sup35 polymorphs.

Discussion
Yeast prion variants are distinguishable based on intensity of the prion phenotype, stability or instability of prion propagation, sensitivity of prion stability to overproduction or deficiency of several chaperones and other cellular components and ability to overcome barriers to transmission between species [30,31,37-41] -or even within species, the last documented here for transmission across the barriers found in wild strains of S. cerevisiae. Yeast prion amyloids are all folded parallel in-register b-sheet structures [21,50,51], but within this architectural restraint, different prion variant structures are proposed to vary in the extent of the b-sheet structure (how much of the N and M domains are in b-sheet), the locations of the folds in the sheets and the association of protofilaments to form fibers.
We find that separation of prion variants based on sensitivity to intra-species barriers cuts across separation based on 'strong' vs 'weak' assessment of strength of prion phenotype. The four transmission variant types derived from the [PSI + ] in strain 779-6A were all strong [PSI + ], like the parent prion. Interestingly, the prions in wild strains were all weak [PSI + ], presumed to arise independently and thus not part of the same 'prion cloud', but fell From each of subclones Y1, Y2 and Y5 from Table 5 were isolated ten subclones, which were then propagated a further .75 generations and clones were amplified and used as cytoduction donors to the three polymorphs. The results from Table 5  After crossing an intraspecies barrier, we find that the [PSI + ref] examined is unstable in its new host, emphasizing the effectiveness of these barriers. We also find that the rare [PSI + ] prions found in wild strains are, in most cases, sensitive to the intraspecies barriers, suggesting that these barriers have evolved to protect yeast from the detrimental effects of this prion.
The [PSI + ] in strain 779-6A, with the reference Sup35p sequence, showed a reproducible strong preference for the reference sequence, transferring only very inefficiently to the D19 or E9 Sup35 backgrounds. However, simple mitotic growth of this strain resulted in the mitotic segregation of at least four variants distinguished by their abilities to cross intraspecies barriers. These variants were stable and reproducible with limited expansion of the corresponding clones, but following many generations of growth, each of those tested gave rise again to the same four general classes of subclones. Prion mutation is well documented in mammals and in yeast under selective conditions [30,46,47,53], and Weissmann's group has suggested that prions resistant to a drug can arise during prion propagation in tissue culture cells in the absence of the drug [54,57]. We observe changes in the predominant prion variant under non-selective conditions in vivo. Selection only happens during the test, when Although multiple de novo prion generation events in forming amyloid in vitro result in multiple prion variants on transfection into yeast, even a [PIN + ] cell generates [PSI + ] clones too rarely to explain our results as de novo prion generation. Rather, mistemplating must be the mechanism of generation of variant diversity that we are observing. Our results imply that there must be a finite rate of amyloid mis-templating that is not due to a mismatch of two prion protein sequences. In spite of extensive purification by mitotic growth and subcloning, we were unable to obtain a prion variant that was completely stable in its transmission pattern to polymorphs. These results are consistent with the 'prion cloud' hypothesis [56,57], in which it is supposed that even a prion variant purified by end-point titration consists of Figure 2. The prion cloud model [56,57] applied to yeast. Segregation of different prion variants on mitotic growth is followed by reemergence of different variants, presumably due to mis-templating. doi:10.1371/journal.pgen.1003257.g002  Table 5. doi:10.1371/journal.pgen.1003257.g001 a major variant as well as an array of minor variants. This production of new prion variants during non-selective growth is analogous to the generation of RNA virus mutants during viral replication (reviewed in ref. [58]), in which a cloud of sequence variants accumulate because of the error-prone nature of RNAdependent RNA polymerases.
The segregation of a mixed prion population could be considered analogous to the segregation of differently marked plasmids with the same replicon. The latter situation has been carefully examined by Novick and Hoppenstadt [59], who find that the fraction of cells remaining with a mixture of plasmids is H = H 0 [(N21)(2N+1)/(2N21)(N+1)] n , where H 0 is the starting fraction of mixed cells, N is the copy number of the plasmid, and n is the number of generations [59]. Random replication of plasmids and equal partition at mitosis is assumed. One result of this treatment is that after N generations, H<0.36 H 0 .
The copy number in the case of yeast prions might be taken as the 'seed number' determined by the methods developed by Cox et al. [60], found to be ,20-120 for the strains examined. The assumption of equipartition is probably not accurate here, since yeast daughter cells are smaller than mother cells [60]. Moreover, the sticky nature of amyloids might suggest that progeny filaments might stick to parent filaments exaggerating this effect. We have propagated our [PSI + ] strains for a number of generations comparable to the presumed copy number, so segregation of different prions is not surprising.
However, we find that even when we have apparently purified a variant, further non-selective growth and subcloning leads to further appearance of the full range of variants among the progeny (Figure 2). This indicates that we are not only observing segregation, but also the (repeated) generation of variants during growth. While varying with respect to transmission, they remain 'strong' variants, suggesting that the structural differences responsible for this transmission barrier differ from those involved in the strong vs. weak differences. King has shown that residues 1-61 are sufficient to propagate strong vs weak prion strains [8,61], but the sequence differences among the Sup35 polymorphs are outside this area, and transmission variants may thus largely differ in the region C-terminal to the 1-61 area, perhaps a region with more variable structure. Other studies have indicated effects of this region on propagation of some prion variants [52], and b-sheet structure of Sup35NM amyloid extends throughout N and even into M [21,22].

Nomenclature
We refer to the standard laboratory yeast sequence [62][63][64] as the 'reference sequence'. Two common sequence polymorphs found within the wild population were used. The first, with deletion of 19 amino acids from residues 66 to 84 and the G162D change, is referred to as D19, and the other includes N109S, G162D, D169E, P186A, T206K, H225D and is denoted E9 [20]. A prion originating with the Sup35p sequence of strain E9, for example, but being propagated in a strain expressing only the reference sequence will be designated [

Scoring the [PSI + ] prion
Sup35p is a subunit of the translation termination complex, and the incorporation of a large proportion of Sup35p into the prion amyloid filaments makes it inactive, resulting in increased readthrough of termination codons. This is measured by read-through of ade2-1, with an ochre termination codon in the middle of the ADE2 gene. In addition to ade2-1, strains carry the SUQ5 weak suppressor mutation, which leaves cells Ade-unless the [PSI + ] prion is also present [2].

Strains, plasmids, and media
The strains used are listed in Table S1. Plasmids used containing reference, D19 or E9 sequences were generated as described [20]. All yeast media and plates contained 20 mM copper sulfate unless noted. Rich and minimal media (YPAD and SD) are as described [65]. Only nutrients required by the strains used in a given experiment were added to minimal plates.

Cytoduction
Cytoplasm may be transferred from one strain to another utilizing the kar1-1 mutation [66], defective for nuclear fusion. Cells fuse, but the nuclei do not fuse, and nuclei separate at the next cell division. However, cytoplasmic mixing has occurred, and so a genetic element (prion or mitochondrial DNA) present in one strain (identified by its nuclear genotype) will be transferred to the other. We use transfer of mitochondrial DNA as a marker of cytoplasmic transfer, and score prion transfer. Reference, D19 or E9 sequence plasmids were transformed into both laboratory strains 4828 and 4830, loss of p1215 (URA3 SUP35C) was selected by growth on 5-fluoroorotic acid media and Ade-transformants were made rhou by growth on YPAD containing 1 mg/ml ethidium bromide. Donor and recipient strains at high density were mixed in water at a ratio of about 5:1, and the mixture was spotted onto a YPAD plate. After 18 hours at room temperature, the mating mix was streaked for single colonies on media selective against growth of the donor strain. Clones are shown to be cytoductants by their growth on glycerol and failure to grow on media selective for diploids. As further tests of a sample confirm, Ade+ cytoductants are judged to have received and propagated [PSI + ].

Subcloning
[PSI + ] Strain 779-6A [48] was streaked to single colonies on K YPD media and twelve colonies were selected, named Y1-Y12. These isolates were streaked to single colonies three additional times, each time selecting just one colony for further propagation. From the third plate a single colony was selected and expanded on K YPD, and cells from this plate were used for cytoduction. From dilution tests there are approximately 2610 7 cells per colony, indicating a total of at least 75 generations of growth of clones Y1-Y12 before cytoduction. Additional subclones were handled in the same manner with only ten colonies selected from the initial K YPD plate. In experiments to rule out selection during stationary growth phase, subclones of Y1 and Y2 were grown in a 125 ml Erlenmyer flask containing 25 ml of liquid YPD medium. When A 600 reached 0.3, the culture was diluted, transferring 1000 cells of each to a fresh flask. These subclones were propagated in exponential phase for 84 generations and were then streaked for single colonies on K YPD plates. After one day of growth on K YPD, 10 subclones were selected for each of Y1 and Y2, expanded and tested for transmission via cytoduction.

Wild [PSI + ] strains
Strains reported to be [PSI + ] [49] were obtained from the UC Davis Department of Viticulture and Enology culture collection. The cultures were first tested to determine if dots were visible using either reference sequence Sup35NM-GFP pDB65 or E9 sequence Sup35NM-GFP pDB81 [20]. Images were obtained with a Nikon Eclipse TE2000-U spinning disc confocal microscope with 1006 NA 1.4 Nikon oil lens with 1.56 magnifier and captured with a Hamamatsu EM-CCD ImagEM digital camera with IPLab version 4.08. Wild strains were sporulated and spores were crossed on rich medium with strain 4972 selecting G418-resistant prototrophs. The diploids formed were again sporulated and tetrads were dissected for each wild strain except for strain 978, whose diploid with 4972 would not sporulate. Ade positive segregants were tested for guanidine curing using two successive streaks on YPAD with 5 mM guanidine. MATa strains were cytoduced into strain 4830 carrying pRS316 (URA3) for selection. Lys2 mutants of MATa strains were selected on plates with DL-aaminoadipic acid as a nitrogen source [67]. Selected strains were retested for Ade positive growth and curing and cytoduced into strain 4828.

Statistical methods
The cytoduction data follows the binomial distribution, because each data point expresses two alternative results, transmission of [PSI + ] or failure of its transmission. However, because of the large number of observations, the results should be approximately normally distributed. We want to calculate the probability that two sets of data are due to chance. Let p 1 and p 2 be the observed proportions of transmission in cytoductions 1 and 2, and n i the number of cytoductants tested in each experiment. Let p = (p 1 n 1 +p 2 n 2 )/(n 1 +n 2 ) be the average of the proportion of transmission in the two experiments. The estimated standard error of the difference between the two proportions is S~sqrt p 1{p ð Þ 1=n 1 ð Þz1=n 2 ½ : The null hypothesis is that cytoductions 1 and 2 are samples from the same population with transmission efficiency p and standard error S. Then the expected proportions are expected to be the same and their difference is expected to be zero. [(p 1 2p 2 )20]/ S = z = the number of standard deviations that the observed difference in proportions differs from the expected difference (0). The frequency of ''z'' being greater or equal to the observed value (assuming the null hypothesis) is obtained from a table of the normal distribution. The calculated ''p values'' are shown in the tables and at appropriate points in the text. Cytoductants examined have been treated as independent since the chance that they represent sister cells is close to zero. This is because cytoductant mixtures were incubated at 20C where the cells divide slowly and because only about 30 cells were examined from several million in the zygote mixture on each plate. Figure S1 Aggregation of Sup35-GFP in reported wild [