The heritable variation that drives new forms and functions is generally ascribed to mutations in the genetic code. We previously proposed an entirely different pathway for creating heritable phenotypic diversity1
, through which the inheritance of new traits can precede the genetic changes that ultimately hardwire them. The mechanism for this seemingly paradoxical flow of information resides in epigenetic switches encoded entirely by self-perpetuating changes in protein structure, known as prions.
The best studied prion is the yeast translation-termination factor Sup35. Like other prions, Sup35 carries a prion-determining domain (PrD) that is dispensable for the protein’s normal function. This PrD occasionally adopts an amyloid conformation. When it does, that amyloid perpetuates itself by templating the same conformation to the PrDs of other Sup35 molecules. This sequesters most Sup35 into insoluble fibres2
. The ensuing reduction in translation-termination activity increases stop codon read-through, producing a variety of new traits that depend upon previously cryptic genetic variation.
Just as the mitotic apparatus ensures inheritance of chromosomally determined traits, the prion-partitioning function provided by Hsp1043,4
ensures inheritance of prion phenotypes. Hsp104 is a molecular machine that severs prion fibres, allowing replicating prion templates to be faithfully inherited by daughter cells. The prion formed by Sup35 is known as [PSI+
], brackets denoting its cytoplasmic inheritance and capital letters its dominant phenotypes.
Cells expressing Sup35 in the non-prion [psi−
] state spontaneously switch to [PSI+
] at a frequency of ~1 in 106
). We have proposed that [PSI+
] provides a beneficial “bet-hedging” mechanism to enhance survival in the face of fluctuating environments: By the time a yeast colony has reached appreciable size, a few [PSI+
] cells will have appeared, expressing heritable new traits. If the trait is detrimental, only a few individuals in a large population will be lost. However, if it is advantageous, those few cells might ensure survival under conditions when the population would otherwise perish. [PSI+
] is also lost sporadically. This guarantees that [psi−
] cells will arise in [PSI+
] colonies, providing a complementary survival advantage.
A particularly attractive feature of this mechanism is that it provides immediate access to genetically complex traits1,7
. Regions downstream of stop codons frequently accumulate genetic variation. [PSI+
]-mediated read-through allows this previously cryptic variation to have biological consequences at multiple loci simultaneously. The complex traits produced by this prion would be less likely to evolve if the individually contributing mutations had to be selected for as they arose. In the long run, reduced translational fidelity should be detrimental. However, advantageous phenotypes initially dependent on [PSI+
] might be assimilated by various means7
, allowing the prion to be lost and the trait maintained.
Several lines of evidence support this hypothesis. First, mathematically, even an infrequent selective advantage for [PSI+
] would be sufficient to maintain Sup35’s prion switching capacity8,9
. Second, the sequence of Sup35’s PrD is highly divergent but has retained, for at least 500 million years, two unusual features that regulate bi-stable inheritance of prion and non-prion phenotypes. An extreme amino-acid bias in one segment drives the PrD into a self-templating prion amyloid, whereas an immediately adjacent charged segment stabilizes it in the soluble non-prion state. Third, the rates at which cells switch into and out of the prion state increases when cells are not well-suited to their environments and new phenotypes have a better chance of being beneficial10
. Increased switching is a direct consequence of the effects that diverse environmental stresses have on protein folding and homeostasis11,12
and also fulfils a critical theoretical prediction for such an evolvability function6,13
In addition to Sup35, at least two-dozen other proteins can form prions that are transmitted through the prion-partitioning activity of Hsp10414,15
in laboratory yeast. These prions are strikingly enriched in transcription factors and RNA-processing proteins that are well situated to transduce genetic variation into phenotypic effect. They too, therefore, might enable the inheritance of diverse biological traits, enhancing survival in fluctuating environments. However, as attractive as such ideas may be16
, and as intensely studied as yeast prions have been, their proposed adaptive value remains highly controversial17–19
. Indeed, prions are often categorized as rare “diseases” of yeast or mere artefacts of laboratory cultivation. A key argument is that [PSI+
] and other prions with phenotypic consequences have not been found in wild strains, despite attempts to find them. Here we establish the natural biological importance of prions through biochemical and biological analyses of hundreds of wild strains.
[PSI+] occurs in wild strains
To search for [PSI+
] in wild strains we took advantage of the unusual stability of prion amyloid assemblies in ionic detergents, which enables their identification by semi-denaturing detergent-agarose gel electrophoresis (SDD-AGE)4
. We modified the technique to enable high throughput detection. Ultimately we analysed 690 yeast strains from diverse ecological niches (Supplementary Table 1
). Amyloid polymers of Sup35 were present in ten (, Supplementary Table 1
, Supplementary Fig. 1
Figure 1 Identification and verification of prions in wild yeast. a, A representative SDD-AGE blot of wild yeast isolates probed with antibodies against Sup35 and Rnq1. Amyloid polymers produce characteristic smears. SDD-AGE does not reliably detect monomeric (more ...)
To ensure that these strains were not simply derived from a recent, prion-containing common ancestor, we sequenced the genomes of two. Over 25,000 single nucleotide polymorphisms established their independent origins (Supplementary Fig. 2
). We also sequenced the SUP35
gene in several of the strains, which established that they, too, had independent origins (Supplementary Table 2
Sup35 amyloid inheritance in wild strains depends on the Hsp104 prion-partitioning factor
Do the Sup35 amyloids in these strains represent true prions? Owing to its central role in the inheritance of prion templates, even transient inhibition of Hsp104’s protein remodelling activity heritably “cures” cells of their prion elements. We inhibited Hsp104 function by growth on medium containing low concentrations of guanidine hydrochloride (GdHCl), which selectively inhibits its ATPase activity, and then plated cells back to media without GdHCl. In all cases this eliminated the amyloid (black arrows, and Supplementary Fig. 1B
). To ensure that curing was not due to an off-target effect of GdHCl we employed a genetic approach – transiently expressing a dominant negative Hsp104 variant3
, on a plasmid marked with antibiotic resistance. (Wild strains contain no auxotrophies.) This also cured cells of the amyloid, confirming their prion-based inheritance (black arrows, and Supplementary Fig. 1B–C
Laboratory culture is not responsible
Might these prions have arisen simply as an artefact of laboratory culture conditions? In archiving wild strains, great care is taken to maintain their wild character (personal communication, Linda Bisson). To directly determine if the conditions employed might have inadvertently selected for the de novo
appearance of [PSI+
], we compared growth of the archived strains with their cured derivatives on all of the relevant media. No growth advantage was found for [PSI+
] on any of these media (YPD, yeast potato dextrose, YM broth, FM broth, wort agar, or Wallerstein nutrient agar) in any of the strains, and it was sometimes detrimental (Supplementary Table 3
). However, in strain #5672 [PSI+
] produced an extreme selective advantage on synthetic grape must, a medium that recapitulates conditions of early wine fermentation20
. This suggests the prion was advantageous in the strain’s natural ecological niche (Supplementary Fig. 3
). In any case, this and several other experiments (Supplementary Information
), indicated that prions harboured by the wild strains almost certainly originated in the yeasts’ natural environments.
Wild [PSI+] is associated with [RNQ+]
In the laboratory, Sup35’s switch to the prion state strongly depends upon the prion-enabling factor [RNQ+
] is itself a prion formed by the Rnq1 protein22
] is the only prion previously known to exist in wild strains18,23,24
. We screened our collection for [RNQ+
] amyloids, finding them in 43 strains (). These, too, depended on the prion-partitioning factor Hsp104 (arrows, and Supplementary Fig. 1
). The correlation between [RNQ+
] and [PSI+
< 0.0001, Fisher’s exact test) was striking: all the [PSI+
] strains contained [RNQ
]. This strongly indicates that [RNQ+
] acts as a prion-inducing factor for [PSI+
] in nature.
[PSI+] transforms natural genetic variation
Do wild prions generate phenotypic diversity from otherwise-cryptic natural variation? We compared growth of the wild [PSI+] strains with that of their cured derivatives in four carbon sources, under osmotic, oxidative, pH, or ethanol stresses, and in the presence of antifungal drugs or DNA damaging agents. We also assessed their ability to invade the growth substratum.
Prion curing produced many phenotypic changes that varied with the genetic background (). For example, strain UCD#824, isolated from white wine, was resistant to acidic conditions and to fluconazole (). UCD#939, isolated from Lambrusco grapes, was resistant to the DNA damaging agent 4-nitroquinoline 1-oxide (4-NQO) (). These beneficial phenotypes were greatly reduced by prion curing. UCD#978, isolated from Beaujolais wine, was sensitive to the oxidative stressor tBOOH () and became more resistant on curing. This same strain normally penetrated the agar surface, but this ability was lost after prion curing (). Thus, in UCD#978 the prion produced a trade-off, creating traits that were potentially detrimental or beneficial, depending on the circumstances.
Figure 2 Prion-contingent phenotypes of [PSI+] isolates. a, Wild [PSI+] strains show diverse phenotypes that are eliminated by transient Hsp104 inhibition. Growth curves for wild strains and cured derivatives in the indicated selective condition are in closed (more ...)
Figure 3 Genetic assimilation of the [PSI+]-dependent adhesive phenotype in meiotic progeny of UCD978. a, [PSI+] allows the wine yeast UCD#978 to adhere to agar surfaces. Adhesion is eliminated by GdHCl or by b, Expression of a non-aggregating version of Sup35, (more ...)
GdHCl and Hsp104DN
cures cells of other Hsp104-dependent prions in addition to [PSI+
. To determine if such curable phenotypes arose from [PSI+
] itself, we transformed the ten strains with a plasmid expressing a Sup35 variant lacking the PrD (Sup35ΔPrD). This protein is immune to [PSI+
]-mediated sequestration and restores normal translation termination in [PSI+
] cells without altering other prions14
. In most cases Sup35ΔPrD produced the same changes as curing with GdHCl and Hsp104DN
(, , see SI for discussion). Thus, most of the original traits were due to [PSI+
Fixation of a [PSI+]-dependent phenotype
When laboratory strains of diverse backgrounds are crossed and sporulated, meiotic re-assortment of the genetic variants they contain can lead to the fixation of a prion trait1,7
. That is, while the trait initially depends on the prion, it can become prion-independent. Might this mechanism allow wild strains to drive prion-dependent traits to fixation? Wild yeasts frequently harbour considerable heterozygosity and sequencing had revealed that the [PSI+
] strain UCD#978 was highly polymorphic. We asked if simple re-assortment of these polymorphisms could fix a [PSI+
Thirty random haploid progeny of UCD#978 were tested for agar adhesion before and after curing. Five retained [PSI+]-dependent adhesion; twenty were no longer adhesive, with or without [PSI+]; five remained adhesive even after [PSI+] curing (). Given the frequency of fixation, it probably required the re-assortment of a few different polymorphisms. But clearly, the naturally occurring genetic variation present in this strain was alone sufficient to fix this trait.
[MOT3+] occurs in wild strains
Nearly two-dozen proteins with prion-forming capacity have recently been discovered in yeast (reviewed in 15
). Serendipitously, an endogenous hexa-histidine motif in one, the transcriptional repressor Mot3, permits detection on SDD-AGE immunoblots14
. Sixteen yeast proteins contain a hexa-histidine motif, but only Mot3 has a prion-like sequence14
. We found [MOT3+
] amyloids in six of the 96 diverse strains we tested (Supplementary Table 1
To determine if wild [MOT3+] prions produced potentially adaptive phenotypes, we first took advantage of Mot3’s known function as a transcriptional repressor of genes involved in cell wall production. We tested wild [MOT3+] strains for resistance to the cell wall toxin calcofluor white. Strain Y-35, isolated from holly berries, was highly resistant to calcofluor. Resistance was heritably reduced by GdHCl treatment, and this treatment also eliminated [MOT3+] amyloids ().
Figure 4 Prions of the cell wall-remodelling transcription factor, Mot3, have diverse phenotypic consequences in wild strains. a, Strain Y-35 is resistant to the cell wall targeting drug, calcofluor white, but resistance is strongly reduced by passage on GdHCl. (more ...)
As a transcriptional repressor, when Mot3 switches into or out of its prion form it has the potential to broadly transform information flow. We next screened wild [MOT3+] strains and their cured derivatives against the same growth conditions used for the wild [PSI+] strains. Many phenotypes were altered by prion curing. For example, [MOT3+] NCYC#3311, a Finnish soil isolate, was resistant to acidic conditions. [MOT3+] Y-1537, isolated from grape must, was resistant to fluconazole. Both traits were virtually eliminated by curing with GdHCl ().
To determine if the traits were [MOT3+]-dependent, we expressed a Mot3 protein lacking the PrD (Mot3ΔPrD) that is immune to prion sequestration but retains normal transcriptional function (Halfmann et al. in preparation). Analogous to Sup35ΔPrD, this eliminates [MOT3+] phenotypes without affecting other prions. NCYC#3311 lost acid resistance and Y-1537 lost fluconazole resistance with this plasmid, but not with plasmids expressing the full-length protein (). These phenotypes were, therefore, [MOT3+]-dependent. More broadly, the divergent consequences of this prion in different strains establish that, like [PSI+], [MOT3+] provides phenotypic diversity by altering the manifestation of natural genetic variation.
Wild strains harbour additional prions
How commonly do wild strains harbour prions that can create such heritable phenotypic diversity? Lacking means of detecting them by SDD-AGE, we employed a phenotypic approach: we measured the growth of wild strains before and after GdHCl curing, across the same conditions used for [PSI+
] and [MOT3+
] (Supplementary Fig. 4
). To ensure that any such phenotypes did not arise from de novo
mutations, we compared four colonies of each wild strain with four cured derivatives (in total testing 5520 isolates across 12 conditions).
Remarkably, over a third of the original wild strains (255) had phenotypes that differed in the same way between all four parental wild strains and all four cured derivatives. Moreover, nearly half of the growth properties conferred by these GdHCl-curable heritable elements were beneficial (Supplementary Table 1
). The wild strain collection was biased toward wine isolates derived from natural fermentations. But it also contained many samples from beer, soil, fruit, infected human patients, and commercial strains recently subject to man-made selective pressures to enhance properties for baking and brewing. Curable phenotypes, both beneficial and detrimental, occurred in yeasts from all of these niches. Even among the limited number of conditions tested here, prion curing had mixed phenotypic consequences in 15% of the strains. Thus, like [PSI+
] and [MOT3+
], these prions created different trade-offs – traits that were beneficial or detrimental depending on circumstances.
To test whether the altered phenotypes arose from prion-mediated protein templating or from some unknown (yet somehow heritable and reproducible) effect of transient GdHCl exposure, we investigated 25 randomly chosen strains more rigorously. Transient expression of Hsp104DN
phenocopied the effects of GdHCl curing in 22 of the 25 strains (Supplementary Table 1
), establishing their dependence on this prion-partitioning factor.
Another signature of prions is transmission to other cells via cytoplasmic transfer (cytoduction). Do the curable phenotypic elements of wild yeasts share this property? Due to complexities in working with wild yeast, we used a derivative of W303, a common laboratory strain, as a universal “recipient” for cytoplasmic transfer. Because prion-dependent traits can differ with genetic background, we chose wild strains with multiple curable traits as donors for these unknown prions, to increase the likelihood that transfer could be scored phenotypically. The South African wine strain WE372 had two traits that were heritably lost by prion curing (, arrows): unusually robust growth on rich medium and poor growth at pH 9. The clinical isolate YJM428 had three such traits (, arrows): robust growth in sodium chloride and 4-NQO, but poor growth on maltose.
Figure 5 The curable Hsp104-dependent epigenetic elements in wild yeast can be cytoplasmically transferred. a, Growth of wild strains WE372, YJM428, and their cured derivatives in selective conditions. Original isolates of each strain are shown in grey bars and (more ...)
After crossing donor and recipient strains to produce heterokaryons, we selected buds that bore only the nucleus of the W303 recipient but had received cytoplasm from the wild donor ( top; see Methods for details). We tested twelve such cytoductants from each mating to determine if they had inherited stable new traits from the cytoplasmic transfer.
Poor growth at pH 9 was not transferred, but robust growth on rich medium was transferred from WE372 donors to all twelve W303 recipients (red arrows, ). NaCl resistance was not transferred, but both enhanced growth on 4-NQO and poor growth on YP-maltose was transferred from YJM428 donors to all twelve W303 recipients (red arrows, ). The fidelity of the transferred traits established that they were not due to rare chromosome transfers that can occur in such crosses. The lack of transfer for some traits suggests that, as for [PSI+] and [MOT3+], traits produced by these unknown heritable cytoplasmic factors depend upon the genetic background. All transferred traits were curable by passage on GdHCl (black arrows, ), strongly indicating they were due to prions. Excluding the possibility that these traits were due to mitochondrial DNA transfer, we repeated the entire experiment with WE372 and YJM428 donors that had been cured of prions before heterokaryon formation (, bottom). None acquired the new phenotypes ().
Prions alter the relationship between genotype and phenotype in wild strains
How significantly do the prions of wild yeasts alter the phenotypic manifestation of genetic diversity? We examined the relationship between genotype and phenotype in the 21 strains in our collection whose genomes had been fully sequenced. As previously reported25,26
, the Spearman’s correlation (rho) between genotypes and phenotypes is typically on the order of 0.3 to 0.4 for wild yeast (in our strains and conditions, rho = 0.39; P
= 3.5 × 10−15
). Prion loss made the correlation between genotype and phenotype weaker (Spearman’s rho = 0.27; P
= 1.5 × 10−7
). This finding was robust to random permutations of the data (P
=0.0001) and was clear even when [PSI+
] and [MOT3+
] strains were removed from the analysis. Thus, the prions these wild strains harbour broadly interface with polymorphisms in their genomes to influence the relationship between genotype and phenotype.