This protocol describes a high-throughput methodology to functionally screen for protein-based inheritance in S. cerevisiae.
The encoding of biological information that is accessible to future generations is generally achieved via changes to the DNA sequence. Long-lived inheritance encoded in protein conformation (rather than sequence) has long been viewed as paradigm-shifting but rare. The best characterized examples of such epigenetic elements are prions, which possess a self-assembling behavior that can drive the heritable manifestation of new phenotypes. Many archetypal prions display a striking N/Q-rich sequence bias and assemble into an amyloid fold. These unusual features have informed most screening efforts to identify new prion proteins. However, at least three known prions (including the founding prion, PrPSc) do not harbor these biochemical characteristics. We therefore developed an alternative method to probe the scope of protein-based inheritance based on a property of mass action: the transient overexpression of prion proteins increases the frequency at which they acquire a self-templating conformation. This paper describes a method for analyzing the capacity of the yeast ORFeome to elicit protein-based inheritance. Using this strategy, we previously found that >1% of yeast proteins could fuel the emergence of biological traits that were long-lived, stable, and arose more frequently than genetic mutation. This approach can be employed in high throughput across entire ORFeomes or as a targeted screening paradigm for specific genetic networks or environmental stimuli. Just as forward genetic screens define numerous developmental and signaling pathways, these techniques provide a methodology to investigate the influence of protein-based inheritance in biological processes.
Biological systems frequently experience transient fluctuations in protein abundance. Whether these have a lasting impact in shaping the phenotype of an organism or of future generations remains unclear. The best-known instances of this biology involve a rare class of proteins, prions, which drive the emergence of heritable traits without genome modification. Instead, these proteinaceous and defectious particles transmit phenotypes via self-perpetuating changes to protein conformation1,2. This type of inheritance was discovered as the cause of the unusual inheritance patterns of a devastating neurodegenerative disease. However, studies in organisms ranging from fungi to mammals3,4,5,6,7,8,9,10 have since revealed that prion-like elements can confer adaptive value. Nonetheless, prions have been viewed as a fascinating but rare biological oddity.
This prevailing wisdom is in part held because the characterization of protein-based inheritance has long been restricted by a small set of examples. Recent systematic screening efforts have widened this picture significantly by identifying several new bona fide prions11 and almost two dozen protein domains12 with the capacity to fuel prion-like conformational conversion. However, because these approaches have generally focused on strong amino acid sequence biases, the prions that have been discovered share the biochemical properties of the founding yeast prions [PSI+]13,14, [URE3]15, and [RNQ+]11,16. These include: 1) modular domains that are rich in long polymeric stretches of asparagine (N) and glutamine (Q), 2) assembly into an amyloid [PRION+] conformation17,18,19, and 3) complete reliance on disaggregase Hsp104 function for faithful propagation from mother to daughter13,20,21. Indeed, many bona fide prions, including [GAR+], [Het-s], and even the original prion (PrPSc), would be missed under such stringent criteria. Perhaps more importantly, they would be unable to capture any novel mechanisms of protein-based inheritance22. Thus, the true biological breadth of such phenomena may be far more common in nature than previously assumed.
To investigate this question, a high-throughput, proteome-wide strategy was employed. A hallmark of all prions, including PrPSc, [GAR+], and [Het-s], is that the transient overexpression of the causal proteins strongly increases the rate of prion acquisition15,23,24,25,26. We took advantage of this feature to systematically ask, across the entire yeast ORFeome, if stable protein-based, epigenetic states could be initiated by transiently inducing the overexpression of individual proteins. It is well known that protein overexpression can alter phenotypes27. However, prion proteins are unusual because their temporary overproduction produces a change in phenotype that is heritable for many hundreds of generations after the initial overexpression. We previously took advantage of this feature, as well as the unusual inheritance patterns of protein-based genetic elements, to identify dozens of proteins that are capable of heritably re-wiring phenotypic landscapes without altering the genome28. Although some identified proteins were previously known as prions, most were not, underscoring the power of this approach to uncover new forms of protein-based inheritance.
1. Initial Overexpression
2. Tests for Prion-like Inheritance
Protein overexpression is known to dramatically alter cellular phenotypes27. Indeed, with an initial screening approach, hundreds of new phenotypes were reproducibly recovered from the overexpression of clones from the yeast ORFeome using just ten stressors. However, the assays described above allow for the assessment of whether cells retain any long-term stable phenotypes following this overexpression. One protein capable of encoding such a state is Psp1. Psp1 is a protein of unknown function that can suppress mutants in the replication machinery (e.g., in pol alpha34). Overexpression of the PSP1 ORF modulated cellular phenotypes in a variety of stressors. Strikingly, one phenotype (resistance to manganese chloride) was maintained for hundreds of generations in cells long after overexpression had ceased (i.e., progeny who had never experienced Psp1 induction directly, but whose ancestors had) (Figure 1A). The half-life of Psp1 is ~5 h35, making it extremely unlikely that the resistance phenotype is due to the propagation of long-lived, stable Psp1 protein from mothers to their daughters over this timescale (as the original protein would be degraded within a handful of generations).
Because the MnCl2 resistance phenotype arose at a high frequency and in multiple independent transient overexpression experiments, it is extremely unlikely that this phenotype was due to de novo mutation. A possible explanation for the heritable nature of this phenotype is the induction of a self-templating element (i.e., a prion). Indeed, the Psp1 amino acid content harbors some small stretches of asparagine and glutamine, reminiscent of canonical prions such as [PSI+], which are driven by proteins that contain long stretches of these amino acids. Predictive algorithms36 score the N-terminus of Psp1 as moderately "prion-like" (Figure 1B, Rank 173 in the yeast proteome). However, unlike canonical prion proteins, Psp1 protein from cells harboring the stable phenotypic state did not form amyloid fibers, as judged by semi-denaturing agarose gel electrophoresis28,37. To determine whether Psp1 formed a bona fide prion, we tested whether the pattern of inheritance of the Psp1-induced phenotypic state was consistent with hallmarks of prion biology: reliance on protein homeostasis machinery for faithful propagation and non-Mendelian inheritance in meiosis. Finally, a "gold standard" test of transmission was performed using only assembled protein.
Unlike mutations or other forms of epigenetic inheritance, prions are uniquely reliant on the function of molecular chaperones and other arms of the protein homeostasis network that regulate protein folding in the cell. Most canonical prions require the Hsp104 disaggregase to act on the amyloid fibers that they adopt. This process generates heritable "seeds" that transmit the prion conformation from mothers to daughters13. One notable exception to this requirement is the [ISP+] prion, which is curable via Hsp104 inhibition but does not require its disaggregase activity for propagation38. Furthermore, we and others have recently reported many examples of prions that are Hsp104-independent and are instead regulated by other molecular chaperones (primarily Hsp70 and, in rare cases, Hsp90)25,28. We first tested whether the inheritance of the Psp1-induced MnCl2 resistance depended on Hsp104. Cells harboring this state were passaged 3 times on rich medium (YPD) supplemented with a low dose of guanidine hydrochloride (an Hsp104 inhibitor). We then returned the cells to standard rich medium to restore full Hsp104 function. This regimen did not affect the heritability of the Psp1-induced MnCl2 resistance, establishing that the cellular state was Hsp104-independent (Figure 2).
Next, a crossing approach was employed to test whether the Psp1-induced phenotypic state depended on Hsp70. The MnCl2 resistant cells were mated to an Hsp70-deficient strain, with genetic deletions in two of the four yeast Hsp70 paralogs (ssa1Δ ssa2Δ). Diploids were selected and then sporulated to generate haploid spores. We tested for the maintenance of the Psp1-dependent state and found that the Psp1-induced MnCl2 resistance phenotype had been completely eliminated in spores harboring the ssa1Δ ssa2Δ double deletion (Figure 2). Thus, the Psp1-induced phenotypic state requires the activity of Hsp70 to pass from one generation to the next.
Another defining characteristic of prions is a non-Mendelian pattern of inheritance. Traits encoded by mutations in DNA segregate 2:2 in meiotic crosses: half of the progeny inherit one parental allele and other half inherit the other parental allele. The inheritance of prions, in contrast, does not depend upon changes in DNA, but rather transmissible changes in protein conformation. Thus, they are inherited by most, if not all, meiotic progeny in genetic crosses (3:1 or 4:0 inheritance)39. We tested the inheritance pattern of the Psp1-induced MnCl2 resistance by crossing strains harboring the state to naïve controls of the opposite mating type. The resulting diploids were sporulated and tested for the presence of the Psp1 MnCl2 resistance phenotype within spores derived from the same tetrads. All spores from 2 separate tetrads in which one parent harbored the Psp1-dependent epigenetic state inherited the corresponding MnCl2 resistance phenotype (Figure 2). In contrast, no cells from naïve control crosses displayed the phenotype. These data establish that the Psp1-induced phenotypic state is not encoded by a DNA-based mutation. However, it is important to note that while the Psp1 state is not encoded in DNA, the state does require the continual presence of the PSP1 gene. Crossing the Psp1-dependent phenotypic state into a genetic background lacking the PSP1 gene (psp1Δ) heritably eliminated the MnCl2 resistance phenotype (Figure 2). Thus, overexpression of Psp1 does not simply orchestrate the formation of a highly stable feedback loop that functions autonomously of Psp1 once initiated.
The very definition of protein-based inheritance is the transmission of stable traits to progeny using only assembled protein as the heritable material. To carry out such "protein transformations," a method that is agnostic with respect to protein stoichiometry, modification, or interaction was used40,41. Separate cultures of cells harboring the Psp1-induced phenotypic state and naïve controls were grown to mid-exponential phase, and their cell walls were lysed to generate donor spheroplasts. These spheroplasts were then sonicated and spun down to pellet the resulting cell debris. Finally, the remaining lysates were subjected to over-digestion with both RNase and DNase to eradicate nucleic acid (to aid steps later in the protocol, the biotinylated DNase enzyme was subsequently removed via affinity purification).
These nucleic acid-depleted lysates were used to transform naïve cells. To aid in the uptake of heritable protein "seeds," the cell walls of the naïve recipients were enzymatically digested prior to transformation. A URA3-marked carrier plasmid with the protein donor lysates was also included to enable the selection of cells that were competent for the uptake of foreign material. The transformants were plated on medium lacking uracil, resistant colonies were selected, and the URA3-marked carrier plasmid was removed by growth on 5-FOA. The resistance of the resulting colonies to MnCl2 was then examined. Strikingly, over half of the transformants tested harbored heritable MnCl2 resistance (Figure 3). In contrast, no cells transformed with naïve lysates displayed any such phenotype. Importantly, these efficiencies were gleaned from cellular lysates expressing endogenous levels of Psp1 protein (~340 molecules per cell). This is remarkably more efficient than the transformation of other yeast prions, such as [PSI+]. These results establish that Psp1 is capable of forming a protein-based heritable element-a "prion"-henceforth known as [PSP1+]. Furthermore, the protein transformation technique described here is sufficient for screening for non-amyloid, Hsp104-independent, protein-based inheritance in a high-throughput manner.
Figure 1. Transient overexpression of PSP1 drives heritable MnCl2 resistance. (A) Growth of uninduced ([psp1–]) and Psp1-induced ([PSP1+]) strains in SD-CSM with 10 mM MnCl2. The error bars represent the standard error of the mean from three biological replicates. (B) Left: The Psp1 primary amino acid sequence scores as highly prion-like in prion-predicting algorithms (PLAAC). Right: Representative PLAAC analysis for most inducing proteins recovered in our screen predicted virtually no significant prion-like sequence features. Please click here to view a larger version of this figure.
Figure 2. Psp1 chaperone dependence and pattern of inheritance. The growth of strains in SD-CSM with 10 mM MnCl2, normalized to a corresponding naïve [psp1–] control. The inhibition of Hsp104 (GdnHCl "cured") does not impair Psp1-dependent MnCl2 resistance, whereas the removal of the Hsp70 chaperone (ssa1Δ ssa2Δ) and the PSP1 gene (psp1Δ) heritably eliminates this phenotype. All spores from single tetrads of crosses between strains harboring the Psp1-dependent phenotypic state and naïve strains display MnCl2 resistance, indicating non-Mendelian (4:0) inheritance of the phenotype (n=2 tetrads). The error bars represent the standard error of the mean from three biological replicates. Please click here to view a larger version of this figure.
Figure 3. Transmission of Psp1-dependent phenotypic state via heritable protein "seeds." Rates of infectivity for protein transformation comparing mock ([psp1–] lysates) and [PSP1+]- harboring lysates as the transmissible material. Infectivity is calculated as the percent of transmission divided by the amount of seeded protein used. Over 53% of naïve cells transformed with seeded [PSP1+] received the corresponding MnCl2 resistance phenotype, whereas no cells transformed with [psp1–] displayed MnCl2 resistance (p=7.4 x 10-4 by Student's t-test). The infectivity rate of the previously characterized yeast prion [PSI+] is displayed via the dotted horizontal line40. Please click here to view a larger version of this figure.
The first yeast prions were identified by their unusual phenotypes and perplexing patterns of inheritance. The characteristics of these prions were then used to build algorithms and computational tools to screen for additional prion proteins. The method described here, in contrast, is experimental and relies on transient overexpression to create a lasting change-a stable state-encoded in protein conformation. However, if the efficiency of "seeding" prion assembly by overexpression for any given protein is very low, that protein will continually emerge as a false negative in overexpression screens of this type. One such modification to correct this would be to use a 2-micron plasmid for protein overexpression in future experiments. Finally, each induced prion has its own unique set of growth phenotypes and will not be apparent in every condition assayed. Thus, the number of different conditions and doses tested limits the number of hits.
Importantly, not all types of protein-based inheritance will be equally recovered using this method. Proteins that cannot be overexpressed efficiently or without toxicity will obviously be continually missed. Mitotically unstable elements, such as "mnemons," would never propagate to daughters following the initial overexpression22. In contrast, other types of long-lived bistable switches could theoretically be induced via transient overexpression42,43. However, these states are typically not dependent on protein homeostatic machinery or transmissible via "seeded" proteins. In addition, prions that rely on other chaperones (outside of Hsp70 and Hsp104) or additional arms of the protein homeostasis network for propagation would fail the chaperone-dependency assays described here. Finally, a low abundance proteins that also form amyloid might have infectivity rates below the limit of detection in the protein transformation setup.
This protocol describes a technique for inducing stable protein-based epigenetic states via protein overexpression, as well as further downstream steps to validate whether each induced epigenetic state is a bona fide prion. The example presented in this paper, Psp1, is an example of a protein that displays a "prion-like" amino acid bias and could theoretically be recovered using previously developed bioinformatic algorithms. However, the inability of Psp1 to form amyloid and its unusual chaperone dependence (Hsp104) would have quickly disqualified it from further analyses and thus eliminated it from prion consideration. However, the screening techniques presented in this paper are agnostic to these assumptions and instead focus on the underlying patterns of inheritance and the sufficiency of protein alone to transmit the corresponding phenotypes. Indeed, the vast majority of protein-based inheritance recovered with this method was devoid of N/Q-rich sequence bias.
This method was used to probe the entire yeast ORFeome for its capacity to elicit protein-based inheritance in an unbiased fashion using only a small number of stressors (25 mM cadmium chloride, 1 mM cobalt chloride, 2 mM copper sulfate, 1 mM diamide, 0.2 mM fluconazole, 50 mM hydroxyurea, 20 mM manganese chloride, 0.75 mM paraquat, 50 mM radicicol, 80 J/m2 UV-irradiation, and 10 mM zinc sulfate). However, this approach could easily be modified to screen genetic networks or specific cellular responses in a more targeted manner. For example, functionally related proteins or all proteins regulated in a discrete signaling network could be induced via transient overexpression and screened with stressors related to their biological function. In contrast, a larger set of proteins could be screened with a more comprehensive set of stressors to investigate whether specific cellular responses have naturally evolved to harbor prion switches. Finally, although we have conducted these studies in yeast, many aspects of the experiments (e.g., transient protein expression, chaperone "curing," etc.) could be generalized to other model systems in the future. For example, mammalian tissue culture is amenable to overexpression via plasmid-based systems, and fluorescence foci could be also be used as a readout for heritable self-assembly, as described previously28. Furthermore, protein-coding sequences from other organisms could be expressed in yeast and tested for their ability to elicit prion-like inheritance using the methods described here.
The authors have nothing to disclose.
We thank Sohini Chakrabortee, Sandra Jones, David Garcia, Bhupinder Bhullar, Amelia Chang, Richard She, and Susan Lindquist for their assistance in developing the assays used in this paper, as well as the reviewers for their thoughtful comments.
Guanidine hydrochloride | Sigma | Cat#G3272-25G | Chemical |
Manganese chloride | Sigma | Cat#M8054-100G | Chemical |
Ethidium bromide | Sigma | E1510 | Chemical |
5-Fluoroorotic Acid | Sigma | Cat#F5013-50MG | Chemical |
BY4741 MATa (his3Δ1 leu2Δ0 LYS2 met15Δ0 ura3Δ0) | Winston et al., 1995; Brachmann et al., 1998 | N/A | Yeast strain |
BY4741 MATα (his3Δ1 leu2Δ0 lys2Δ0 MET15 ura3Δ0) | Winston et al., 1995; Brachmann et al., 1998 | N/A | Yeast strain |
Hsp70 (K69M) | Jarosz et al., 2014b | N/A | Plasmid |
FLEXGene library | Hu et al., 2007 | N/A | Plasmid library |
Dextrose (glucose) | Fisher Scientific | D16-3 | Media component |
Raffinose | Sigma | R0250-25G | Media component |
Galactose | Fisher Scientific | BP656-500 | Media component |
CSM | Sunrise Science | 1001-100 | Media component |
CSM-URA | Sunrise Science | 1004-100 | Media component |
CSM-LYS | Sunrise Science | 1032-100 | Media component |
CSM-MET | Sunrise Science | 1019-100 | Media component |
CSM-LYS-MET | Sunrise Science | 1035-100 | Media component |
yeast extract | Fisher Scientific | BP1422-2 | Media component |
peptone | Research Products International | P20240-5000 | Media component |
bacto-peptone | BD | 211677 | Media component |
glycerol | EMD Millipore | GX0185-2 | Media component |
yeast nitrogen base w/o amino acids | BD | 291920 | Media component |
agar | IBI Scientific | IB49172 | Media component |
Adenine sulfate | Sigma | A3159-25G | Media component |
Potassium acetate | Sigma | P1190-500G | Media component |
Uracil | Sigma | U0750-100G | Media component |
Histidine | Sigma | H8000-100G | Media component |
Leucine | Sigma | L8000-25G | Media component |
Lysine | Sigma | L5501-25G | Media component |
RNase I | Thermo Fisher Scientific | EN0601 | Enzyme |
biotinylated DNase | Thermo Fisher Scientific | AM1906 | Enzyme |
zymolyase 100T (yeast lytic enzyme) | Sunrise Science | N0766555 | Enzyme |
Microplate reader | BioTek | Synergy H1 | Equipment |
Microplate stacker | BioTek | BioStack3 | Equipment |
Plate filler | BiotTek | EL406 | Equipment |
Liquid handling robot | Beckman Coulter | Biomek FX | Equipment |