Creating a Protein-Based Element of Inheritance

See allHide authors and affiliations

Science  28 Jan 2000:
Vol. 287, Issue 5453, pp. 661-664
DOI: 10.1126/science.287.5453.661


Proteins capable of self-perpetuating changes in conformation and function (known as prions) can serve as genetic elements. To test whether novel prions could be created by recombinant methods, a yeast prion determinant was fused to the rat glucocorticoid receptor. The fusion protein existed in different heritable functional states, switched between states at a low spontaneous rate, and could be induced to switch by experimental manipulations. The complete change in phenotype achieved by transferring a prion determinant from one protein to another confirms the protein-only nature of prion inheritance and establishes a mechanism for engineering heritable changes in phenotype that should be broadly applicable.

Two genetic elements in Saccharomyces cerevisiae, [PSI +] and [URE3], are widely believed to transmit phenotypes through proteins with self-perpetuating changes in conformation, rather than through altered nucleic acids (1). These elements are called yeast prions because of conceptual similarities between their modes of transmission and that postulated for the infectious agent in mammalian prion diseases (2). The yeast proteins, however, are unrelated to the mammalian prion protein and to each other. Moreover, they usually do not kill the organism, but produce cytoplasmically transmitted, heritable changes in phenotype (3,4). For [PSI +], the protein determinant is Sup35, a translation termination factor. In [psi ] cells, Sup35 is soluble and functional (5, 6). In [PSI +] cells, most Sup35 is insoluble and nonfunctional, causing a change in translation fidelity (5,6). This phenotype is heritable because Sup35 protein in the [PSI +] state influences new Sup35 protein to adopt the same state and passes from mother cell to daughter to perpetuate the cycle of conversion (6–8). [PSI +] is, however, metastable: [PSI +] cells occasionally give rise to [psi ] cells and vice versa (3), as the [PSI+ ] conformation is lost or gained.

Sup35 has three distinct regions (9). The NH2-terminus (N) plays a critical role in Sup35's self-perpetuating change in state (10–13). The middle region (M) provides a solubilizing and/or spacing function (14). The COOH-terminus (C) provides translation-termination activity (10, 15). To test whether other proteins can be made to undergo a prion-like change in state, we fused N and M to a steroid hormone–regulated transcription factor, the rat glucocorticoid receptor (GR), and to a constitutive variant (GR526) (16) that lacks the heat shock protein 90 (Hsp90) and hormone-binding domain (Fig. 1A) (17). The fusion did not block GR's inherent transcriptional activity: when NMGR and GR were expressed in [psi ] cells over a broad range of induction levels, the activity of NMGR was in each case similar to that of GR (Fig. 1B, left). NMGR526 and GR526 also had similar activities in [psi ] cells (18).

Figure 1

(A) Derivation of NMGR and NMGR526 expression construct from Sup35 (N, M, and C regions) and GR (AD, activation domain; DBD, DNA-binding domain; and LBD, ligand-binding domain) (9, 16). NMGR was subcloned into both a singlecopy, copper-inducible vector (26) (pCUP1-NMGR) and a multicopy (2μ) constitutive vector (16) (pG1-NMGR). NMGR526 was subcloned into the constitutive 2μ vector (pG1-NMGR526). (B) GR and NMGR activities assessed (27) by activation of a β-galactosidase reporter construct, pL2/GZ (16). Three individual transformants were picked for each construct and duplicate samples were measured. Left: copper was added at the indicated concentrations to [psi ] (74-D694) cultures to induce GR (open bar) and NMGR (filled bar). Right: GR and NMGR activities in isogenic [psi ] and [PSI +] cells (74-D694) at the same level of induction (100 μM CuSO4).

Three lines of evidence indicate that NM fusion proteins can interact with endogenous Sup35 to undergo prion-like changes in state. First, in contrast with [psi ] cells, in [PSI +] cells, newly synthesized NMGR had much lower activity than GR (Fig. 1B, right). Immunoblotting demonstrated that this was not due to a reduction in NMGR expression (18). Second, transient expression of NMGR, but not of GR, induced new heritable [PSI +] elements in [psi ] cells (Table 1). The NMGR526 variant also induced [PSI +] elements, but not if the protein carried a small deletion of residues 22 through 69 in N (Δ22–69) (Table 1), which blocks [PSI +] induction by Sup35 (10). Third, NMGR and NMGR526 exhibited the same unusual pattern of plasmid incompatibility as Sup35 (10, 13). High-copy Sup35 plasmids can- not be transformed into [PSI +] cells because excessive Sup35 aggregation inhibits translational termination so severely that cells die. Cells are immune to the toxicity of the plasmid when the genomic copy of SUP35 has its N and M regions deleted (ΔNMSUP35) (10, 13). Conversely, plasmids are not toxic if they carry the Δ22–69 deletion. When GR and NMGR were transferred to a high-copy vector with the yeast glyceraldehyde-3-phosphate dehydrogenase (GPD) promoter, a strong, constitutive promoter, transformants were readily obtained in [PSI +] cells with GR and GR526, but not with NMGR and NMGR526 (Table 1). This transformation restriction was eliminated if the plasmid carried theΔ22–69 mutation or if the genome contained theΔNMSUP35 mutation (Table 1).

Table 1

[PSI +] induction and transformation incompatibility. Isogenic [psi ], [PSI +] and ΔNMSUP35 (74-D694) were transformed with single-copy, copper-inducible plasmids or high-copy constitutive plasmids encoding GR, NMGR, or their COOH-terminal deletion derivatives with or without the Δ22-69 mutation. To test for [PSI +] induction, [psi ] transformants were grown in selective liquid media in the presence of 100 μM copper (for inducible plasmids) or without copper (constitutive plasmids) overnight. Five μl of the culture was then plated onto synthetic medium without adenine [SC (–ade)], which is selective for [PSI +] cells but not for the plasmid. Representative [PSI +] colonies were tested for curing by GdHCl. The (+) indicates at least 20 [PSI +] colonies were obtained; (−) indicates no [PSI +] colonies were obtained. To assess transformation competence, cells were plated onto medium selective for the plasmid. Each transformation was repeated at least three times. (+), 20 to 200 transformants per plate; (−), no transformants obtained. As expected, [PSI +] cells induced from [psi ] cells by high-copy, constitutive plasmids died during continuous selection for the plasmid and could only be recovered if selection for the plasmid was dropped.

View this table:

To determine if NMGR and NMGR526 can undergo self-perpetuating prion-like changes in activity on their own, we transformed the constitutive expression plasmids into aΔNMSUP35 strain (19). In addition to providing a test for the autonomy of NMGR conversions, this eliminated the potentially complicating toxicity of endogenous Sup35 inactivation. No changes in translational fidelity were detected in this background in our experiments.

Transformants were screened for different GR activity states by using the convenient blue/white colony color assay for a GR-regulated β-galactosidase reporter. With GR and GR526, all transformants yielded blue colonies and gave rise only to blue colonies on restreaking. That is, the proteins were transcriptionally active and remained so. However, with NMGR and NMGR526, white colonies were occasionally obtained (<1%) (18). On restreaking, blue NMGR and NMGR526 colonies primarily gave rise to blue colonies and white to white (Fig. 2), but each occasionally gave rise to colonies of the other color, and these in turn occasionally switched back. Thus, both NMGR and NMGR526transformants exhibited heritable but metastable changes in their ability to activate the β-galactosidase reporter gene. Because NMGR526 assays did not require the addition of hormone, further work concentrated on this construct.

Figure 2

NMGR and NMGR526 exist in distinct stable functional states. (A) Blue (active) and white (inactive) NMGR isolates (28) of 74-D694ΔNMSUP35 cells containing pL2/GZ, and pG1-NMGR (middle and right) or pG1-GR (left) were streaked onto selective medium (SC –trp, –leu). After 2 days of growth they were replicated onto selective medium containing 80 μg/ml of X-GAL. (B) Visualization of blue and white NMGR526 cells. Shown are 74-D694 ΔNMSUP35 cells containing pG1 vector (left) or pG1-NMGR526 (middle and right). (C) Immunoblot analysis. Equal quantities of total cell proteins prepared from white and blue NMGR526 by ethanol lysis were separated by SDS-PAGE and immunoblotted with BuGR-2, a monoclonal antibody to GR (29).

White colonies produced the same level of transcription factor as blue colonies (Fig. 2C), indicating that differences in GR activity were not due to differences in expression level. To determine if white colonies resulted from changes in the genome or reporter plasmid, white cells were streaked to media that caused them to lose the NMGR526 plasmid. When they were retransformed with GR526, all new transformants were blue (18), indicating that both the reporter plasmid and the genome were fully capable of supporting β-galactosidase activity. When the same cells were transformed with NMGR526, most colonies were blue, but a few were white. That is, in the absence of NMGR526, all previous determinants of the GR activity state were lost and new NMGR526 transformants were as likely to be blue or white as were the initial transformants. We conclude that white NMGR526 colonies produced NMGR526 protein with a different heritable functional state than that of blue colonies, and the maintenance of this state depended upon continuous production of NMGR526.

Next, we studied whether the inactive state of NMGR526 was “infectious.” Cells expressing inactive NMGR526 (white colonies) were mated to cells expressing active NMGR526 or NMGR (blue colonies) from plasmids with different selectable markers. The resulting colonies were white and remained white even after the original NMGR526 plasmid was lost (18). We conclude that heritable differences in the activity of NM fusion proteins are due to their ability to form “infectious” protein-based genetic elements, or prions (1).

Other prions switch between active and inactive states spontaneously (as does NMGR) but can be induced to switch at much higher rates by experimental manipulation. NMGR functional states can be manipulated similarly. When cells from blue (active) NMGR526 colonies were transformed with an expression plasmid for the N domain and grown overnight in liquid media, ∼10 to 20% yielded white colonies, and these remained white when the plasmid was lost (18). Virtually all colonies derived from vector-only transformants were blue. Thus, as for Sup35, transient overexpression of N can heritably convert NMGR526 to the inactive state. To determine if inactive NMGR526 could be induced to switch to the active state, we employed the protein denaturant guanidine hydrochloride (GdHCl) and the protein remodeling factor Hsp104, which can induce such changes in Sup35 (20,21). When white colonies expressing NMGR526 were replica-plated to medium containing 5 mM GdHCl, most became blue (Fig. 3A). On restreaking in the absence of GdHCl, they remained blue. Similarly, white NMGR526isolates transformed with an Hsp104 over-expression plasmid but not with the vector alone changed to blue (Fig. 3B) and remained blue when the Hsp104 plasmid was lost (22). As with the original transformants, blue colonies derived from GdHCl and Hsp104 treatments produced mostly blue colonies on restreaking but occasionally gave rise to white ones, indicating that the heritable changes they induced in NMGR526 activity were reversible. The effects of Hsp104 on NMGR526 also demonstrate that the critical determinant for Hsp104 interaction is within the NM domain of Sup35.

Figure 3

Reactivation of NMGR526 by GdHCl and Hsp104. (A) Two white isolates (1, 2) of NMGR526 and a vector control (c) in 74-D694ΔNMSUP35 containing pL2/GZ were streaked onto a SC (–trp, –leu) plate. After incubation for 2 days at 30°C, colonies were directly replicated onto either SC (–trp,–leu, +X-GAL) (left panel) or first replicated to SC (–trp, –leu, +5 mM GdHCl), and then replicated onto SC (–trp, –leu, +X-GAL) (right panel). (B) White isolates of NMGR526 in 74-D694ΔNMSUP35 were transformed either with p2HG-104, a 2μ Hsp104 expression plasmid, or p2HG, the vector alone. Three transformants from each construct were randomly chosen and streaked on SC (–trp, –leu) then replica plated onto X-GAL indicator plates after 2 days.

Our data clearly show that a prion-determining function is modular and transferable. When the prion determinant of Sup35 was deleted from its normal location and transferred to a completely unrelated protein, a new prion was created that maintained the epigenetic characteristics of the original prion and was subject to the same types of experimental manipulations (e.g., induction and curing). However, the phenotype changed completely, from one affecting the fidelity of protein synthesis to one affecting the transcription of a glucocorticoid-regulated promoter. Thus, the functional domain of the protein attached to NM is indeed the determinant of phenotype, affirming the protein-only nature of yeast prion inheritance.

Given the heterologous nature of the proteins employed here, it seems likely that this method for creating protein-based elements of inheritance will be broadly applicable. Of course, there will be exceptions. Indeed, fusing the prion-determining domain of Ure2 to β-galactosidase (23) and the NM regions of Sup35 to firefly luciferase (24) caused no change in β-galactosidase or luciferase activity. Even if the prion domains of these fusion proteins were to switch states, the functional domains might remain active because their substrates and products are freely diffusible. Expression levels would also impose another limitation: GR is expressed inefficiently in yeast; we were therefore only able to obtain heritable changes in activity with a high-copy vector. Nevertheless, the list of potential targets for prion manipulations is very large. It includes any protein that must be targeted to a particular location or assembled into complexes with other proteins in order to function. Here, it worked equally well with a transcription factor that requires hormone for activation and constantly cycles through interactions with the multicomponent Hsp90 chaperone machinery (GR) and a constitutive Hsp90-independent variant (GR526). Preliminary results with a bacterial luciferase fusion suggest soluble enzymes might also be susceptible if substrates access or folding pathways can be altered by the prion domain (24). Protein-based genetic elements provide a method for engineering changes in phenotype that is fundamentally different from that of changes induced through the alteration of nucleic acids. Phenotypes switch spontaneously at rates that can be regulated by modifying the sequence of the prion determinant (25). Phenotypes can be induced to switch in opposite directions by transient over-expression of the prion determining domain or protein remodeling factors. Once switches occur they are stably inherited from generation to generation, but yet again, can be reversible by transient inducing stimuli. Perhaps most uniquely, prions provide a general mechanism for engineering loss-of-function phenotypes that are both dominant and infectious. Moreover, because they spread from one protein to others that share the same determinant, they can simultaneously inactivate diverse proteins in a single step.

  • * To whom correspondence should be addressed. E-mail: S-Lindquist{at}


View Abstract

Navigate This Article