|Home | About | Journals | Submit | Contact Us | Français|
Most prions (infectious proteins) are self-propagating amyloids (filamentous protein multimers), and have been found in both mammals and fungal species. The prions [URE3] and [PSI+] of yeast are disease agents of Saccharomyces cerevisiae while [Het-s] of Podospora anserina may serve a normal cellular function. The parallel in-register beta-sheet structure shown by prion amyloids makes possible a templating action at the end of filaments which explains the faithful transmission of variant differences in these molecules. This property of self-reproduction, in turn, allows these proteins to act as de facto genes, encoding heritable information.
The term “prion” means “infectious protein”, without the need for an accompanying nucleic acid, a concept first suggested by Griffith (1) to explain the uniformly fatal transmissible spongiform encephalopathies (TSEs) of mammals. He suggested that an altered oligomeric protein form could catalyze the conversion of the normal form of the protein into the same altered form, and this proposed ‘protein-only’ process could explain the extraordinary radiation-resistance of infectivity (2). Griffith's abnormal oligomer has been realized as ‘amyloid’, the basis of most known prions (see below). Amyloid is defined as a filamentous form of protein with a cross-beta sheet structure (the beta strands are largely perpendicular to the long axis of the filaments). In its amyloid form proteins are usually more protease-resistant than in the normal form, and they show special staining with the dyes Congo Red and Thioflavin T. These self-propagating amyloids, which are the focus of most activity in the prion field, are a model for the many common human amyloidoses that are apparently not infectious, such as Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, type 2 diabetes and many others. As Griffith suggested, there are many possible mechanisms for an infectious protein and an enzyme whose active form is necessary for the activation of its own inactive precursor is another which has been realized in at least one case.
Here we review particularly the biology and structural aspects of amyloid-based prions. The most surprising aspect of prions is that proteins can be genes (just as nucleic acids can be enzymes). We particularly focus on the way in which the structures of prion amyloids may be able to explain their biological role as genes.
Accumulated evidence supports the protein-only (prion) explanation of the TSEs, although some skeptics remain. The PrP protein is the major component of the TSE infectious agent (3). Its coding gene, called Sinc in mice (4,5) determines the TSE incubation period and deletion of Sinc makes mice immune to TSE infection (6). The disease isoform of PrP, a particulate protease-resistant form called PrP-res or PrPSc, can seed the conversion of the normal protease-sensitive form, called PrP-sen or PrPC, into the disease form in vitro (7), providing a biochemical basis for the propagation process. Very low levels of infection have been obtained from amyloid of recombinant PrP in one study (8), but repeated cycling of the in vitro seeding reaction can produce much higher infectivity titers (9). In the presence of non-specific RNA, spontaneous generation of highly infectious PrP amyloid from purified PrPC can be achieved (10).
The long-mysterious non-chromosomal genes of S. cerevisiae, [PSI+] (11,12) and [URE3] (13,14), had not been identified with any nucleic acid replicon. Three genetic criteria were proposed to distinguish nucleic acid replicons from prions (15): a) if a prion can be cured, it can arise again in the cured strain because the protein able to change to an infectious form is still present; b) overproduction of a protein capable of becoming a prion should increase the frequency of the prion arising de novo; c) if the prion phenotype is due to the absence of the normal form, then the phenotype of mutation of the gene for the protein should be the same or similar to that of the prion as both are deficient in the normal form. None of these three properties is expected for nucleic acid replicons and each is expected for a prion (15). All three were satisfied for [URE3] as a prion of Ure2p and for [PSI+] as a prion of Sup35p (15)(Fig. 1A). It was noted that although overproduction of Sup35p increased the frequency of [PSI+] arising in some strains, this was not the case in others(16). Genetic analysis showed that a non-chromosomal genetic element, called [PIN+], for [PSI+]-inducibility, was present in the former strains (16). [PIN+] is now known to be a self-propagating amyloid form of Rnq1p (17), a protein rich in N and Q residues whose gene deletion produces no phenotype (18). Among Q/N rich proteins whose overproduction could mimic [PIN+] is Swi1p, subunit of a chromatin remodeling factor (17). Recently it has been shown that Swi1p itself can be a prion with a phenotype due to deficiency of the normal form of the protein (19).
Ure2p is a negative regulatory factor in nitrogen catabolism, shutting off the transcription of genes encoding the enzymes and transporters needed for the utilization of poor nitrogen sources when a good nitrogen source is available (reviewed in (20)). A ure2 mutant or the presence of the [URE3] prion results in inappropriate expression of a multitude of nitrogen catabolism genes in the presence of ammonia, a good nitrogen source (21), including DAL5, the allantoate permease (22). Dal5p can also transport ureidosuccinate, an intermediate in uracil biosynthesis, whose import can be used to reflect the inactivity of Ure2p resulting from its conversion to amyloid.
Sup35p is a subunit of the translation termination factor (23,24), and its mutation or inactivity because of the prion change results in increased read-through of translation termination codons, so nonsense suppression is used as an assay for [PSI+].
A colony of a filamentous fungus is not a pile of cells, but is a syncytium of interconnected cells. The HET-s protein of the filamentous fungus Podospora anserina is involved in heterokaryon incompatibility. When two colonies of a filamentous fungus grow toward each other, they do a trial fusion of cellular processes, and test identity at about a dozen polymorphic loci (called het for heterokaryon) scattered about the genome (25). If the two colonies are identical at all of these loci, fusion of the two colonies proceeds and they can share nutrients, cytoplasm and even nuclei. A difference at even a single locus results in death of the few fused cells and establishment of a barrier to futher fusion. One of these loci, het-s, has two alleles, het-s and het-S. Remarkably, proper incompatibility of fusion of het-s and het-S strains requires that the HET-s protein (encoded by het-s) be in the amyloid (prion) form (26,27).
Amyloid is the basis for most prions, but the [β] prion of S. cerevisiae is simply the active form of a self-activating enzyme, vacuolar protease B (28). Protease B (Prb1p) is made as an inactive precursor that is normally cleaved and activated by protease A (Pep4p) (29), but in a pep4 mutant, some transient self-activation can be observed (30). If the level of Prb1p is derepressed by growth on a non-fermentable carbon source, then this self activation continues, and has all the properties of a prion (28). Cells harboring inactive protease B can be infected by cytosol with active protease B. This illustrates that self-propagation and infectivity need not be based solely on amyloid. The ‘crippled growth’ non-mendelian genetic element, [C], of Podospora may similarly be a prion based on MAP kinase self-activation (31).
Ure2p, Sup35p and HET-s have structurally distinct terminal prion domains, and [PIN+] can be propagated by a restricted part of Rnq1p. Ure2p1-65 is necessary and sufficient to transmit the [URE3] prion (32,33), and constitutes the core of the amyloid formed in vitro by the full-length Ure2p (34,35)(Fig. 1B). Amyloid of recombinant Ure2p1-65 infects yeast with [URE3](36). Similarly, Sup35p1-123 is necessary and sufficient to propagate the [PSI+] prion (37) and forms amyloid (38) that is infectious, transmitting [PSI+] (39,40). Neither ‘prion domain’ is necessary for the normal function of the molecule, but as discussed below, both aid those functions. Ure2p1-65 appears to be unstructured in native Ure2p (41), but acquires β-sheet structure on conversion to the prion (amyloid) form (34,42). [PIN+] can be propagated by residues 153 - 405 of Rnq1p (43) and amyloid of recombinant Rnq1132-405 is infectious to yeast (44). The prion domain of HET-s is the C-terminal residues 218-289, a region that is largely unstructured in the native form (45). Amyloid of recombinant HET-s218-189 can infect Podospora with the [Het-s] prion (46). The prion domains of Ure2p, Sup35p and Rnq1p are all Q/N rich, although those of HET-s and PrP are not.
There is little doubt that the uniformly lethal mammalian TSEs are diseases. However, the yeast and fungal prions are compatible with normal growth in laboratory conditions, so the question of whether yeast prions are adaptive or detrimental can be an issue. Sup35p, a subunit of the translation termination factor, is essential for translation of all mRNAs, and would thus seem a poor choice for regulation of gene expression by simple inactivation in a messsage non-specific way. Ure2p negatively regulates transcription of genes encoding transporters and enzymes for using poor nitrogen sources, turning off such genes when a good nitrogen source is available. Ure2p activity is already regulated (reviewed in (20)), so sequestering Ure2p would seem a poor strategy as it actually prevents regulation. Deleting Rnq1p has no phenotype (18), so one cannot predict whether [PIN+] could be adaptive. [Het-s] is necessary for heterokaryon incompatibility in Podospora (26), so one might assume it is an advantage to its host (47).
How can one judge whether a prion is a help or a hindrance? If a prion gave its host a general growth advantage, this could argue that it is adaptive. But if it produces a growth advantage or disadvantage, depending on conditions, one would have to know which conditions are well represented in the wild. Moreover, yeast spend most of their time in stationary phase, waiting for nutrients to appear (otherwise the earth would soon be one big yeast colony!), so survival in stationary phase may be more important than growth rate. In fact, more conditions favor growth of [psi-] over [PSI+] than the contrary (48), suggesting that [PSI+] is a disadvantage, but there is no information on how the [psi-] - advantage or the [PSI+] - advantage conditions are represented in the S. cerevisiae natural environmental niche.
The presence of the prion domain of Ure2p and Sup35p in many species (e.g. (49)) has been interpreted to mean that prion formation is conserved and therefore beneficial to the host (Actually, many Sup35's do not have a putative prion domain: Human – NP_ 002085; Rice - NP_001052341; mosquito - XP_320105). This reasoning is appropriate if the prion domain has no non-prion function. In fact, the prion domain of Ure2p is important for the function of the protein in nitrogen regulation, both stabilizing the protein and aiding its interaction with other factors of nitrogen regulation (50). Deletion of the Sup35p prion domain is lethal in combination with certain mutants in Sup45p, the other subunit of the translation termination factor (51). This implies that there is some function normally carried out by either Sup45p or by the Sup35p prion domain, and if both are defective, the cells cannot grow. Moreover, cells lacking the Sup35p prion domain show a variety of growth phenotypes distinct from those shown by isogenic SUP35 wild type [psi-] or [PSI+] strains (48). This clearly implies that there are non-prion functions of the Sup35p prion domain.
Beyond these considerations, although the Ure2p of S. paradoxus has a Q/N-rich N-teriminal extension quite similar to that of the cerevisiae Ure2p, it cannot become a prion at detectable frequency in its native environment, S. paradoxus (52). This result indicates that the presence of a Q/N-rich domain does not imply prion-forming ability, and that in this species, the ‘prion domain’ is conserved for some non-prion reason.
For an infectious element, such as a plasmid, virus or prion, a general test of whether it is adaptive is available. Even clearly detrimental viruses, bacteria and prions are often widespread in nature, simply because they are infectious. Their spread outpaces the damage they do to their hosts. Certainly an infectious element that is an advantage to its host must be easily found in wild strains. The mitochondrion is an obvious example. An infectious element that is not found in wild strains must, on the net, be detrimental (53). A survey of 70 wild Saccharomyces strains did indeed show that each of the mildly detrimental yeast nucleic acid viruses and plasmids were found in a portion of the strains (Table 1), but neither [URE3] nor [PSI+] was present in any of the strains (53). This result implies that [URE3] and [PSI+] are detrimental to their host, a conclusion that is independent of what phenotypes they may produce. [PIN+] was found in 11 of the 70 strains, a proportion similar to that of the mildly detrimental plasmid and viruses in the same strains (Table 1). [PIN+] is probably thus mildly detrimental.
It has been noted that the prion domain of Ure2p varies more rapidly than the C-terminal part of the protein (54). Because the prion domains are less critical for function, their variation may be a result of less selection pressure. However, another possible explanation for this phenomenon is provided by Collinge's finding that heterozygosity in the human gene encoding PrP protects against spontaneous or infectious CJD (55), and his suggestion that the corresponding polymorphism was selected for by this protection in an era when cannibalism was more common (56). The variation among Sup35p's of different Saccharomyces species indeed results in a significant barrier to transmission of [PSI+] (57). It is possible that the rapid variation in sequence in the prion domain of Sup35p and Ure2p was selected for by the unfavorable consequences of being infected with a prion from another strain.
[Het-s] is of special interest because the prion appears to be part of a normal fungal mechanism (heterokaryon incompatibility) to limit fusion of cellular processes to genetically identical partners (see above). This system may be a defense against potentially debilitating fungal viruses which are spread by these fusions (hyphal anastomosis). Strikingly, the proper incompatibility of het-s and het-S strains requires that the HET-s protein be in the prion (amyloid) form. This certainly suggests that the [Het-s] prion is functioning for the organism and is not a disease (47). Indeed, 80% of wild het-s isolates carry the [Het-s] prion (58), as one would expect for a prion benefiting the cell.
However, another possible interpretation is based on the finding that mating a female het-s [Het-s] strain (carrying the prion) with a male het-S strain produces meiotic segregants in which het-S meiotic spores are dead (58). This is an example of “meiotic drive”, a phenomenon in which a chromosomal allele promotes its inheritance not by improving the fitness of the organism, but rather by cheating on meiosis. Examples include the t-locus of mice, with meiotic drive allele ‘t’ and normal allele ‘T’. Males that are T/t heterozygotes produce sperm which are mainly ‘t’. However, t/t homozygous males are sterile. Here a disadvantageous allele becomes widespread in nature, representing 10% of alleles, by preventing inheritance of the normal allele. The presence of the [Het-s] prion in spores with the chromosomal het-S allele results in lethality, and loss of the het-S allele (58). This event, rather than the facilitation of heterokaryon incompatibility, may explain why about half of alleles at this locus are het-s and 80% of wild het-s isolates have the [Het-s] prion.
The presence of an imperfect repeat sequence in Sup35p (59) and the partial conservation of Ure2p10-39 among related yeasts (54) suggested that some prion domain sequences are important for prion formation or propagation. However, when Ure2p1-89 and Sup351-114 were randomly shuffled (keeping their amino acid content unchanged), each of the five shuffled variants in both cases were able to form prions (60,61). This proved that the oligopeptide repeats and conserved sequences are not needed for prion formation. There is, however, a length requirement (61), so that deletion mutants may alter prion formation not because of the specific sequences deleted, but because they make the domain shorter.
How can the sequence independence of prion formation be reconciled with the strict sequence-dependence of prion propagation (65)? Amyloids are known with parallel in-register structure and with antiparallel structure. While β-helix amyloids have been proposed, and parallel out-of-register amyloids are possible, none have been found as yet. In an antiparallel β-sheet, a β-helix or an out-of-register parallel β-sheet, non-identical amino acid residues will be apposed in adjacent protein chains. The sequence specificity of prion propagation could only be maintained if there were some required relation between paired residues (e.g., + with - charge, or large residue with small residue, or hydrophobic with hydrophobic, or hydrophilic with hydrophilic). Shuffling the sequence would certainly disrupt this relation. However, a parallel in-register structure having identical residues paired between chains would be less sensitive to shuffling since identical residues could still pair in the shuffled sequence (65).
The filamentous structure of amyloids (Fig. 2A) makes it impossible to analyze them using X-ray crystallography or solution NMR. X-ray crystallography is limited to structures of microcrystals formed by 4 to 7 residue peptides from various amyloid-forming proteins; these have an array of structures including parallel and antiparallel architectures (66-68). Solid-state NMR of specifically or generally labeled proteins has been used extensively to examine amyloid filament structures (69-72)(reviewed in (73)). Chemical shift values (the resonant frequency) for carbonyl-13C or α-13C residues reflect their secondary structure. Further, the distance between a 13C and the nearest neighbor 13C can be measured by a dipolar recoupling experiment, in which the rate of signal decay varies as 1/r3. For example, the Ure2p prion domain labeled with Leu-1-13C at each of the three Leu residues (positions 12, 16 and 81) shows a rate of signal decay indicating that each carbonyl C is about 0.5 nm from the nearest neighbor 13C, a result hard to explain by any structure other than a parallel in-register β-sheet (42) (Fig. 2B). Solid-state NMR of amyloid usually produces resonances many fold broader than for solution NMR, in part because of heterogeneity of the structures. However, if the resonances are particularly sharp, as in the case of amyloid of the HET-s prion domain peptide (residues 218-289), detailed structural studies are possible (74-76). This work indicates that direct imperfect repeats in the HET-s prion domain are paired in a parallel orientation. This structure can be viewed as a two-turn β-helix, or, because the paired residues are nearly identical, a pseudo-in-register parallel β-sheet (Fig. 2C).
Sup35NM, including both the prion domain (N= residues 1-123) and the adjacent highly charged domain (M= residues 124-253) has 20 Tyr residues scattered throughout N and none in M. The Tyr-1-13C signal decays rapidly, again indicating ~0.5 nm to the nearest neighbor 13C, and a parallel in-register β-sheet (77). The eight Leu residues include one within the N domain and seven in the M domain. Signal decay of Leu-1-13C Sup35NM is slower than for the Tyr-1-13C labeled sample, but indicative of several Leu residues within the M domain also being in parallel β-sheet structure. This is consistent with evidence that M affects [PSI+] prion propagation (78,79).
Amyloid of Rnq1p153-405 can transmit the [PIN+] trait to yeast cells (44), and preparations labeled with either Tyr-1-13C or Leu-1-13C show signal decay in dipolar recoupling indicating ~0.5 nm distance to the next nearest Tyr-1-13C or Leu-1-13C, respectively (80). Similar experiments with Ala-3-13C show that the structure is not out of register by even a single amino acid residue (80). These results again imply a parallel in-register β-sheet structure.
The relatively narrow filament diameters observed for the yeast prion domain amyloids (18,34,38) implies that these β-sheets must be folded over several times along the long axis (the fiber axis). However, in no case is the location of the folds (turns in each polypeptide) known.
In each case, the amyloids studied include several variants (36,39,40,44), but Weissman's group has found conditions under which the infectious Sup35NM filaments produce preferentially ‘strong [PSI+]’ or preferentially ‘weak [PSI+]’ (40). Particularly promising is King's demonstration of extensive faithful propagation of two variants in vitro using purified [PSI] particles as seeds and a fragment of the Sup35 prion domain fused to GFP as substrate (39).
While finding a parallel in-register β-sheet constrains the structure quite substantially, this is not the same as having an atomic-level structure. Detailed evidence-based models of Aβ1-40 (81) and the diabetes-related amylin 37- residue peptide (82) required residue-by-residue labeling of synthetic peptides, an approach not feasible for the much longer prion domains.
Two prion phenomena are explained by the parallel in-register structure of the amyloid-based yeast prions: a) the heritability of specific prion properties (prion ‘variants’) and b) the Pin phenomenon: the seeding of one prion by another.
It is remarkable that a single protein species can become any of several prion ‘variants’ or ‘strains’, even in genetically identical hosts. This phenomenon was first recognized in mouse scrapie, where clearly distinct incubation periods, disease symptoms and distributions of brain lesions in genetically identical mouse lines left little doubt that the infectious agent had different properties (reviewed in (83)). This phenomenon was actually used as an argument against the prion model since it was considered unlikely that a protein could encode information in this way. However, a difference in protease resistance of PrP between TSE strains was demonstrated (84) making this an important line of evidence for the protein-only (prion) model.
Prion variants are also observed in yeast prions [PSI+] (85), [URE3] (36,86) and [PIN+] (87), where intensity of the phenotype and stability of inheritance of the prion are the variables measured. Propagation (or not) by various Sup35 prion domain mutants can also be used to distinguish variants (88). As in the TSEs, structural differences between the amyloids determining yeast prions have been demonstrated (39,40,89). Deuterium exchange experiments (89), scanning mutagenesis and deletion mutant studies (90) indicate that variants differ in the stability and extent of structures in the prion domain, although it is as yet unclear what is the precise structure of any prion variant.
The solid-state NMR structural studies were done on material that is almost certainly a mixture of prion variants, so the results represent what is common among most or all of them. The parallel in-register structure implies that each molecule, as it joins the end of the filament, could contact the full length of the preceding molecule to join the filament. This allows transmission of the location of turns, the length of loops and any irregularities in the β-strands from one molecule to the next. This templating step, analogous to the templating of complementary DNA strands, is the essential feature for any inheritance phenomenon to be based on amyloid filaments, and is understandable in light of the parallel in-register β-sheet structure.
How can one explain the seeding of one prion by another prion (17)? The fact that [PSI+], [URE3] and [PIN+] each involve a Q/N rich domain of their respective proteins is clearly important. The common parallel in-register structure suggests that it is the misincorporation of a Sup35p or Ure2p molecule at the growing end of an Rnq1p filament that results in the conversion of what began as a [PIN+] filament into a [PSI+] or [URE3] filament. This is, in any case, a rare event, and it does not violate the rule that prion propagation is sequence specific.
While not a focus of this review, the reader should note the prominent role played by chaperones and other cell components in prion propagation, first noted with the discovery that depletion or overproduction of the disaggregating chaperone Hsp104 eliminated the [PSI+] prion (91) (reviewed in (92))(Fig. 3). Hsp70s and Hsp40s also affect the propagation of various yeast and fungal prions (e.g., (93-100). At least one role of chaperones is the scission of amyloid fibers to make new seeds, a process necessary for propagation of prions (101-103). Overproduction of the Hsp40, Ydj1p, may cure [URE3] by binding to the native protein (104), thus reducing incorporation into amyloid.
The yeast and fungal prions are important models for mammalian prions and amyloids. It is likely that with five prions in S. cerevisiae and two in Podospora anserina there are more prions to be discovered. While a start has been made on understanding the structure of the infectious amyloids, for none is an atomic level structure available. Different prion variants have been shown to have different structures, but we do not yet know in detail what those structures are. While many chaperones affect prion generation and propagation, how these effects, and their selectivity for protein sequence and prion variant are determined remains unclear. It is hoped that study of yeast and fungal prions may facilitate development of anti-prion or anti-amyloid treatments (108,109), perhaps along the lines pioneered by Kelly and co-workers for transthyretin amyloid (110).