Anti-prion systems in yeast

Yeast prions have become important models for the study of the basic mechanisms underlying human amyloid diseases. Yeast prions are pathogenic (unlike the [Het-s] prion of Podospora anserina), and most are amyloid-based with the same in-register parallel β-sheet architecture as most of the disease-causing human amyloids studied. Normal yeast cells eliminate the large majority of prion variants arising, and several anti-prion/anti-amyloid systems that eliminate them have been identified. It is likely that mammalian cells also have anti-amyloid systems, which may be useful in the same way humoral, cellular, and innate immune systems are used to treat or prevent bacterial and viral infections.


Herb Tabor, my mentor
I began my scientific career after completing my medical internship when I became a postdoc with Herb and Celia Tabor in 1967 at the National Institutes of Health. I was fortunate to be working with someone who was such an exceptional model scientist. After 3 years as a postdoc with Jerry Hurwitz in New York, I returned to Herb's department in an independent position. Herb had built a department of diverse interests that was unified by daily literature and weekly work seminars, both featuring intense discussion and critique. It was an excellent environment for learning and developing one's own program. Although I became the Lab Chief (NIH's equivalent of the Department Chair) in 1996, my relationship with Herb has remained essentially unchanged for over 50 years (it took me ϳ25 of those years before I switched from "Dr. Tabor" to "Herb"). Herb Tabor is totally devoted to science, has always done experiments himself, is far more concerned about getting the right answer to a problem than in making a splash, and is the least vain of any scientist I have ever met. His total honesty, sincerity, and scientific focus have been a model I have attempted to emulate (though largely without success). His devotion to science did not mean neglect of his family, but rather a total involvement of his family in science. Celia White Tabor worked together with Herb for many years, and all of his wonderful children have had science careers. A large part of Herb's life has been science in the lab, but an equal or maybe even larger part has been scientific publishing. His technologically forward-looking approach and his broadening the scope and enlarging the size of the JBC had an enormous impact.
Among the lessons I have learned from Herb are: • Any scientific area has the potential to be interesting, if studied with enough depth and skill.
• "Depth is better than width." • The word "critical" was used in a positive sense to describe someone who could thoughtfully, fairly, and accurately evaluate a piece of work.
• "Kinetics are meaningless." • "Some scientists may not want to be acknowledged because they do not think much of your work." • "I've told you all I know, maybe more." Although my own work has taken a largely genetic direction, the lessons I learned from Herb are not specific to any area of science, and I know I have benefited greatly from his constant example.
The word "prion," meaning "infectious protein," without any nucleic acid being required for the transmission of the infection, originated in studies of the transmissible spongiform encephalopathies of mammals (1)(2)(3). However, the proof of the existence of a prion had to await the discovery that the nonchromosomal genetic elements, [URE3] and [PSIϩ] of Saccharomyces cerevisiae, are based on altered infectious forms of Ure2p and Sup35p, respectively (4). Both the mammalian transmissible spongiform encephalopathies and the yeast prions are self-propagating amyloids. Amyloid is a linear polymer of a particular protein, a sort of one-dimensional crystal in which the predominant structure is ␤-sheet, with the ␤-strands perpendicular to the long axis of the filaments. Of late, evidence has been found for prion-like behavior, or even frank infection, for such classical amyloidoses as amyloidosis A (5, 6), Alzheimer's disease (7-10), multiple system atrophy (11), amyotrophic lateral sclerosis (12), and type 2 diabetes (13). This developing unification of the human amyloidoses around the prion model enhances the importance of an in-depth study of the prions of S. cerevisiae, with their facile genetic manipulability. Ure2p, a repressor of genes encoding transporters and assimilation enzymes for poor nitrogen sources, is active when a good nitrogen source is available (16), but in [URE3] strains, Ure2p is largely trapped in infectious amyloid filaments (17)(18)(19)(20). The aggregated form is evidently largely inactive as [URE3] cells have a phenotype similar to ure2⌬ cells, the inappropriate derepression of the controlled genes despite the presence of a good nitrogen source. Sup35p is a subunit (with Sup45) of the translational termination factor (21,22). In [PSIϩ] cells, Sup35p is in infectious amyloid aggregates (23)(24)(25)(26)(27)(28)(29), and increased readthrough of stop codons ensues.
A wide array of other prions has now been found in S. cerevisiae and one particularly revealing prion in the filamentous fungus Podospora anserina (30).
[PINϩ] is an amyloid-based prion of Rnq1p, a protein of unknown function (31)(32)(33) (35). Extensive evidence indicates that this stimulation of prion formation occurs by an inefficient form of the same seeding process that is involved in propagation of all of the amyloid-based yeast prions (36). There is clinical and experimental evidence that similar cross-seeding is an important feature of human amyloidosis (37,38). Because nearly all known pathogenic amyloids have a similar architecture (see below), it is likely that this potentiation of formation of one prion/amyloid by another is a general phenomenon.
Prion variants are a feature of all pathogenic prions, whether of animals or yeast (39). A single prion protein with a single sequence can be the basis of a wide array of prion variants (or prion strains), with distinct biological properties and different amyloid conformations (29,40). Each variant is relatively stably propagated, implying that there must be a mechanism by which the amyloid filaments act as a template to force monomers joining the end of the filaments to assume the same conformation as molecules already in the filament. In yeast, prion variants may differ in the intensity of their phenotype (strong versus weak), stability of propagation, ability to propagate in the face of overproduction or deficiency of various chaperones or other cell components, ability to cross interspecies or intraspecies barriers, and other properties. Using a nonselective system, it was shown that the [PSIϩ] prion exists as a "cloud" of prion variants that segregate from each other as cells grow and mutate at some frequency (41), thus establishing the "prion cloud" model (39,42).
Prion domains are the part of the protein that actually forms the amyloid and is roughly the same as the part needed to transmit the prion (17,19,26,43). Both the extent of amyloid structure and the region needed to faithfully propagate the prion vary with the prion variant (29,44,45). The "prion domains" have normal nonprion functions. The Ure2p prion domain is necessary for the stability of the whole molecule against degradation, and thus for the full nitrogen regulation function (46). The prion domain of Sup35p is necessary for general turnover of mRNAs (47), for cytoskeleton-associated translation (48), and for recovery from the stationary phase (49).

Structure of infectious yeast prion amyloids
In an attempt to show that there were sequences in the prion domains of Ure2p and Sup35p needed for prion formation, it was, surprisingly, found that randomly shuffling these domains produced sequences that in all cases were able to form prions (50 -52)! This proved that it was not the sequence but rather the amino acid content of these domains that made them suitable for prion formation, and detailed analysis has revealed which residues favor or impair prion formation (53,54). The sequence independence of prion formation, combined with the wellknown barriers to prion propagation produced by even a single amino acid difference in some cases (55)(56)(57), indicated that the faithful propagation of prion variant/strain information was not based on "complementarity," as for DNA or RNA, but rather a principle of "identity" (52). Any complementarity feature (self-complementarity in this case) would be destroyed by shuffling the sequence. It was realized that an anti-parallel ␤-sheet, a ␤-helix, or an out-of-register parallel ␤-sheet would rely on complementarity between neighboring amino acids in different molecules (52). However, a parallel, in-register ␤-sheet features rows of identical amino acid residues along the long axis of the filament, such as had already been shown for A␤ amyloid (58). Shuffling the sequence would not prevent identical residues from interacting in a parallel in-register structure, only their order would change. For this reason, we proposed that the prion amyloids of Ure2p and Sup35p had an in-register parallel ␤-sheet architecture (52).
Based on extensive studies of highly infectious amyloids of Sup35p, Ure2p, and Rnq1p, using solid-state NMR (59 -63) and EM (particularly mass/length measurements) (64 -66), it is clear that each of these prion amyloids in fact does have a folded, parallel, in-register, ␤-sheet architecture. Indeed, the amyloids formed by the shuffled prion domains also had this architecture (67).

Mechanism of conformational templating
The fact that a single protein sequence can be the basis for an array of prion variants apparently differing only in protein conformation implies that there must be a mechanism for proteins to template their conformation and to transmit this conformation information to other molecules of the same sequence. This is the central mystery of prions. Based on the folded in-register parallel ␤-sheet architecture of yeast prion amyloids, we have suggested that the same forces of favorable interaction between identical side chains that keeps this architecture in-register also ensures that a new molecule joining the end of a filament will adopt the same structure as molecules already in the filament ( Fig. 1) (68, 69). A line of glutamine residues along the long axis of a filament can form a line of hydrogen bonds, but only if the molecules are in-register. Thus, this interaction stabilizes the in-register architecture. The same is true of a line of asparagines, serines, or threonines. A line of identical hydrophobic residues can have a series of stabilizing hydrophobic interactions if the structure is in-register. Charged residues would of course destabilize an in-register structure, and charged residues are rare in the prion domains of Ure2p, Sup35p, and Rnq1p. Furthermore, mutations blocking prion transmission are often uncharged-to-charged amino acid changes (55,56). A molecule of the prion protein joining the end of the filament, to have these favorable side-chain interactions, must have a turn between the ␤-strands at the same location as those molecules already in the filament. This results in the end of the filament acting as a template to transmit conformation information to the new molecule joining the end of the filament (68,69). This templating explains how proteins can act as genes, templating conformation in analogy to the templating of sequence by replicating DNA or RNA (Fig. 1).
Nearly all of the pathogenic amyloids associated with human diseases that have been examined have the same folded in-register parallel ␤-sheet architecture (58, 70 -77) as observed for these yeast prions, with one exception being transthyretin (78 -80). It is likely that templating of conformational information for the human pathologic amyloids proceeds by the same mechanism discussed here.  (82)(83)(84), more rare than the 2-m DNA plasmid that is known to cause a Ͼ1% growth detriment (82,85). These and other results show that no variants of these prions are beneficial on the net (86).

Anti-prion systems in yeast
There is a wide array of cellular components whose overproduction or deficiency leads to the loss of one or another prion (36,87,88). These studies have been important in establishing mechanisms affecting prions, but it is generally unclear which of these effects represent physiologic (or pathophysiologic) responses of cells to prions. It is of particular interest to know whether there are systems that cure yeast prions in a normal cell, without overproduction or deletion of anything, and it is these that we refer to as "anti-prion" systems. As expected for a pathogenic infectious agent, the cell indeed has defenses against yeast prions at several levels. Prion formation is partially blocked; having formed, several systems cure the nascent prions; and other systems prevent prion-based pathology ( Table  1). As with DNA repair systems, these anti-prion systems are each only partially effective.

Ribosome-associated Hsp70s inhibit [PSI؉] formation
Ssb1 and Ssb2 are Hsp70 family members that are associated with the ribosomes and are involved in folding of nascent polypeptides (89,90). For example, in the absence of the Ssb proteins, aggregated proteins accumulate in the cell (91). In ssb1⌬ ssb2⌬ cells, the frequency of [PSIϩ] arising is about 10-fold higher than in a normal cell (92,93). Restoration of Ssb1 does not cure the [PSIϩ] variants arising in the double mutant, indicating that the Ssb proteins largely prevent [PSIϩ] generation, rather than blocking their propagation (92). However, dissociation of Ssb's from ribosomes in cells grown on minimal medium has a [PSIϩ]-curing effect (94).

Hsp40 family member Sis1p prevents toxicity of otherwise mild [PSI؉] prion variants
Although a majority of [PSIϩ] prion variants are highly toxic to cells, the variants usually studied do not produce an obvious growth defect. The studies of wild strains (above) show that even these variants are detrimental in the wild. However, Masison's group (95) has shown that these "mild" [PSIϩ] variants have the potential to prevent cell growth except for

F-box protein Ylr352w/Lug1 prevents toxicity of mild [URE3]
A general screen for genes that are essential specifically in [URE3] cells was carried out using transposon mutagenesis with the Hermes transposon (96). Genes that could be mutated in a [ure-o] cell, but not in a [URE3Ϫ1] strain, were inferred to be either necessary for propagation of [URE3] or for cell growth/survival in the presence of the prion. The most dramatic hit from the screen was YLR352W (renamed LUG1 for "lets [URE3] grow") (96), encoding a substrate-specifying subunit of a cullin-containing multisubunit E3 ubiquitin ligase (97). Other subunits of this complex and subunit-modifying proteins also were detected in the screen (96). Lug1p is one of 20 substrate-specifying subunits with a common sequence called the "F-box" (97-100), but the substrates specified by Lug1p are unknown. The lug1⌬ [URE3] strains could not grow on glycerol media, whereas lug1⌬ [ure-o] cells grew normally. The growth defect was suppressed by overproduction of Hap4p, a transcription factor that stimulates expression of mitochondria-related genes, consistent with the requirement for oxidation for utilization of glycerol (96). The growth defect of lug1⌬ [URE3] strains was also suppressed by mutation of GLN1, encoding glutamine synthase. Like cells carrying [URE3], gln1 mutants are defective in nitrogen catabolite repression, the diminished expression of genes for using poor nitrogen sources when a good nitrogen source is available. This indicates that the inability to utilize glycerol of the lug1 [URE3] strains was not a consequence of failure to control nitrogen catabolism. Growth of lug1⌬ strains on proline, a condition that is known to shut off Ure2p's activity in repressing NCR genes, does not affect cell growth, confirming the conclusion that Ure2p has another function independent of NCR (96). It remains unclear what is the relevant target of Lug1p, but it is clear that this protein prevents the pathologic effects of an otherwise relatively mild variant of [URE3].
The same screen also showed a substantial decrease in detection of mutations in many chaperone genes, including some that had not previously been implicated in [URE3] prion propagation. This effect was observed for HSP82 and HSC82 (encoding Hsp90 paralogs), HSP104 and SSA2 (known to be necessary for [URE3] propagation (101,102)), YDJ1 and CAJ1 (Hsp40s), STI1 (co-chaperone of Hsp90s), FES1 and SSE1 (nucleotide exchange factors for Hsp70s known to affect [URE3] (103)), HSP26 and HSP42 ("small" heat-shock proteins), SSB1 and SSB2 (ribosome-associated Hsp70s), and SSA1 and SBA1. Whether these effects reflect specific toxicities of [URE3] prevented by the chaperones, or a synergistic toxicity of prion-induced stress and stress due to chaperone deficiency, remains to be determined. Nearly all of these chaperones have close human homologs (HSP104 is a notable exception), making understanding this phenomenon of particular interest.

Normal levels of Btn2p and Cur1p cure most [URE3] variants
Overproduction of either of two paralogs, Btn2p or Cur1p, cure [URE3] (104). In the process, Btn2p collects the aggregates of Ure2p at one site in the cell (instead of being scattered about the cytoplasm), and this collection spot co-localizes with a spot of Btn2p (104). Apparently, on cell division, the probability of daughter cells getting any filaments is thereby reduced, producing frequent cured cells (104). Consistent with this "sequestration model," Btn2p can also collect nonprion aggregates of other proteins at a single cellular site (105,106). However, Cur1p did not co-localize with Ure2p aggregates in the course of its curing of [URE3] (104).
To determine whether normal levels of Btn2p or Cur1p have an anti-prion action, a series of [URE3] prion variants was isolated in a btn2⌬ cur1⌬ strain, and the normal levels of each protein were restored by mating or by other means (107). Remarkably, nearly all (ϳ90%) of these [URE3] variants were cured in the WT environment, although most were quite stable in the btn2⌬ cur1⌬ strain. Indeed, the frequency of [URE3] arising spontaneously was elevated about 5-fold in btn2⌬ cur1⌬ strains (107). A propagon-measuring method (108) revealed that variants of [URE3] with high seed number (presumably many amyloid filaments per cell) could only be cured by overproduction of Btn2p or Cur1p, whereas those variants with low seed numbers were cured by normal levels of these proteins (107). Thus, there are far more prions arising than the few that manage to escape the cellular antiprion systems.

Anti-prion protein
Anti-prion protein activity Prion affected Effect Ref.  (110). Btn2p was previously known as a mediator of protein trafficking in the endosome pathway (111,112), and Btn3p, an inhibitor of Btn2p's action in protein sorting, also inhibits its ability to sequester the [URE3] amyloid filaments and cure the prion (113). This is the only connection yet defined between the protein-sorting and aggregate-collecting activities of Btn2p.

Normal levels of the Hsp104 disaggregase cures most [PSI؉] variants
Hsp104 is a disaggregating chaperone that, in concert with Hsp40s and Hsp70s, disaggregates denatured proteins allowing cells to recover from heat shock (114). Hsp104, working with Hsp70s, Hsp40s, and nucleotide exchange factors, cleaves prion amyloid filaments to generate new filament ends and thus new growing points (23,87,(115)(116)(117)(118)121). However, overproduction of Hsp104 cures the [PSIϩ] prion (115) and [URE3] to a minor extent (104). The propagation-promoting activity of Hsp104 involves its removing monomers from the middle of an amyloid filament thereby breaking it into two pieces (119, 120). Inhibition of Hsp104 by millimolar concentrations of guanidine results in the arrest of new filament formation, and eventual curing of the prion in most progeny cells as the filaments are diluted by cell growth (108,(122)(123)(124).
The mechanism of the [PSIϩ]-curing effect of Hsp104 overproduction is still controversial, despite extensive study. One line of evidence shows that Hsp104 has a specific binding site in the M (middle) domain of Sup35p (125) and that excessive Hsp104 blocks access of Hsp70s to the filaments, thereby preventing filament cleavage (126). Another study shows that overproduced Hsp104 results in asymmetric segregation of prion seeds, resulting in frequent loss of the prion (127). A third group proposes that the overproduction curing activity of Hsp104 is a "trimming" activity, removing monomers from the ends (in contrast to cleavage of filaments in the middle) (128).
To determine whether normal levels of Hsp104 cure some [PSIϩ] variants as they arise, we used the same approach as discussed above for Btn2 and Cur1. We used the hsp104 T160M mutant that propagates [PSIϩ] normally but cannot cure [PSIϩ] even if overproduced (129). A series of [PSIϩ] variants isolated in this strain was tested for the ability to propagate in an HSP104 host, and about half were unable to do so (130). The frequency of [PSIϩ] arising spontaneously was elevated over 10-fold in hsp104 T160M hosts. These results indicate that the normal Hsp104 (without overexpression) cures many [PSIϩ] variants as they arise (130).

Inositol pyro-/polyphosphates and [PSI؉] propagation
In a general screen of the knockout collection for mutants defective in anti-[PSIϩ] activity, two isolates were siw14:: kanMX mutants (131), lacking a pyrophosphatase specifically active on 5-pyrophosphoinositol pentakisphosphate, and thus having elevated levels of this compound (Fig. 2) (132). We found that nearly all [PSIϩ] variants require some inositol poly/ pyrophosphate and that 5-pyrophosphoinositol tetrakisphosphate, inositol hexakisphosphate, or 5-pyrophosphoinositol pentakisphosphate could support [PSIϩ] propagation (131). In the absence of the 5-pyrophosphoinositol polyphosphates, elevated levels of the 1-pyrophosphoinositol polyphosphates inhibited [PSIϩ] propagation (131). The target(s) of the inositol poly/pyrophosphates in exerting these effects is at this time unclear. The inositol poly/pyrophosphates have a wide array of activities (133,134), making this a difficult task. Wu et al. (135) prepared affinity resins with the inositol poly/pyrophosphates that support [PSIϩ] and identified yeast proteins that adhere to columns of these resins. Among these were Hsp26p, Sse1p, and Ssb1,2, each of which has been reported to affect [PSIϩ] generation or propagation (92,103,136,137), and they must be considered candidates for mediators of action of the inositol poly/pyrophosphates.

Upf proteins, mediators of nonsense-mediated mRNA decay, cure many [PSI؉] variants
The screen described above also returned deletions of UPF1 and UPF3 (138), genes encoding components of the nonsensemediated mRNA decay apparatus, responsible for elimination of mRNAs carrying a premature stop codon (139). Because these genes also affect the genetic test for [PSIϩ] (readthrough of a premature stop codon), it was shown in several different ways that the effect of the Upf proteins was on prion propagation and not on the assay. The upf mutants have a Ͼ10-fold increased frequency of [PSIϩ] generation, and nearly all of the [PSIϩ] variants produced in a upf mutant are cured by simply replacing the normal amount of the Upf protein (138). The Upf proteins normally form a complex with the translation termination factors Sup35p and Sup45p on the ribosome, and it is argued that this complex stabilizes Sup35p in its normal form preventing conversion to the amyloid form (138). Upf1p, at sub-stoichiometric levels, inhibits amyloid formation in vitro by Sup35p (but not by Ure2p), suggesting that a direct interaction is responsible for blocking prion propagation (138). Examination of a series of UPF1 and UPF2 mutants showed a poor correlation of [PSIϩ]-curing activity (138) with nonsense-mediated decay (NMD) efficiency, RNA helicase, ATPase, or RNA-binding activity, but a better correlation with their reported (140,141) formation of the Upf1-2-3 complex and binding to Sup35p.
It is possible that the binding of Sup35p monomers by the Upf complex competes with their binding to amyloid filaments. Alternatively, the binding of the Upf1-2-3 complex to the ends of Sup35p amyloid filaments may block access of Sup35p monomers to these growth points. We infer that normal proteinprotein interactions naturally compete with and may even reverse abnormal interactions. This may have general implications applicable to treatment of disease.

Study of anti-prion systems reveals an unexpected abundance of prion variants
At this time the various anti-prion systems appear to be specific for a particular prion, although only limited tests have been done. Several systems cure [PSIϩ] variants, and it is not yet known whether they work independently or in concert. It is evident in yeast that prions are arising at a dramatically higher rate than was previously appreciated and that the array of antiprion systems cures most of them. There are also anti-prion systems that prevent prion formation, and others that make some of those prions that do succeed in arising less harmful than they would otherwise be. It is hoped that detailed study of the yeast anti-prion systems will facilitate discovery of homo-logous or analogous mammalian systems active against the many human amyloidoses.