|Home | About | Journals | Submit | Contact Us | Français|
In vivo amyloid formation is a widespread phenomenon in eukaryotes. Self-perpetuating amyloids provide a basis for the infectious or heritable protein isoforms (prions). At least for some proteins, amyloid-forming potential is conserved in evolution despite divergence of the amino acid (aa) sequences. In some cases, prion formation certainly represents a pathological process leading to a disease. However, there are several scenarios in which prions and other amyloids or amyloid-like aggregates are either shown or suspected to perform positive biological functions. Proven examples include self/nonself recognition, stress defense and scaffolding of other (functional) polymers. The role of prion-like phenomena in memory has been hypothesized. As an additional mechanism of heritable change, prion formation may in principle contribute to heritable variability at the population level. Moreover, it is possible that amyloid-based prions represent by-products of the transient feedback regulatory circuits, as normal cellular function of at least some prion proteins is decreased in the prion state.
The central dogma of molecular biology1 provides a specific mechanism for the previously postulated2 template principle in biology. DNA and RNA can be considered as the first order templates, that is, linear or sequence templates, either for each other or for polypeptides. Discovery of infectious proteins (prions),3 and especially of prion mechanism of inheritance4 introduced templates of another type, structural or conformational templates, which could be designated as second order templates.5 According to the current view,4,6,7 the process of propagation of the amyloid-based prions begins with a conformational change in the protein, and is followed by “linear crystallization,” producing amyloid fibers. The new rounds of prion multiplication may be initiated or seeded with preexisting amyloid fragments. Transmission of these fragments in cell divisions results in the inheritance of the prion state in yeast and fungal systems. Physiochemical studies of elementary amyloid particles uncovered the β-rich structure,8–10 in some examples11 held together by the intermolecular parallel β-sheets. Variations of this structure apparently determine patterns of the specific variants, or “strains” of a given prion protein.12
Conformation templating of a yeast prion can be reproduced in vitro,13,14 resulting in generation of infectious prion particles, faithfully reproducing the variant-specific patterns upon transformation into the yeast cells. Yeast and fungal prions known to date are described in detail in other reviews (see refs. 4 and 6). Patterns of the mammalian prion protein PrP have also been reviewed recently (refs. 15 and 16).
Prion propagation is a highly sequence-specific process. Domains forming an axis of the amyloid fiber should be identical to each other at the level comparable to that required for the complementary interaction of nucleic acid sequences. However, aggregating proteins of different sequences can facilitate aggregation of each other in certain assays. For example, de novo appearance of the prion conformation of the yeast protein Sup35, containing a QN-rich prion domain, is facilitated in the presence of the prion isoform of another QN-rich protein, Rnq1,17–19 reflecting the existence of a prion network. Molecular mechanism of this interaction is still unclear, and it does not seem to involve a template-like component.
Amyloid is probably an ancient fold, as almost any protein can form an amyloid in vitro depending on conditions.20,21 Moreover, second order templating is not restricted to prions. Other examples of “structural inheritance” involve inheritance of preformed structures in Protozoa.
As prions and similar phenomena appear to be widespread, the question arises whether these phenomena play a biological role. Two possible models of the biological role of prions were proposed in literature. One model, designated here and further as “prion pathology” model, states that prion (or amyloid) formation is a pathological process, while conservation of amyloid-forming potential in evolution is due to other adaptive functions of prion-forming proteins, which are not necessarily related to prion formation per se (example in ref. 22). Another model, designated here and further as model of “adaptive prionization,” suggests that prion formation by itself could be an adaptive process, so that certain prions are responsible for adaptive traits (example in ref. 23).
Examples of the “prion diseases” are well known and include various infectious neurodegenerative diseases in mammals.15,16 According to the “protein only” concept, which is now accepted by the majority of experts, the PrP protein in its prion form (PrPSc) is the sole component of a “transmissible particle” that is responsible for the genesis and transmission of a disease. Usually, there is a correlation between the disease and cerebral accumulation of PrP.3,24 The properties of PrP are very similar to those seen in various noninfectious amyloidoses and neural inclusion disorders, a large and heterogeneous group including more than 20 human diseases, among them Alzheimer's, Huntington's and Parkinson's diseases,25 resulting from conversion of certain proteins or their fragments from the normally soluble form to insoluble fibrils or plaques.
Although protein-destabilizing mutations can confer the ability to form amyloids in vivo even to such commonly known proteins as lysozyme,26 usually disease-related aggregation depends on the presence of the specific elements of the primary structure. One feature frequently associated with aggregation is the presence of regions within proteins that comprise a single homopolymeric tract of a particular amino acid and are called homopeptide repeats, or SSR (single sequence repeats).27 It has been shown that uncontrolled genetic expansions of SSR regions lead to the development of some neurodegenerative disorders, for example Huntington's disease, associated with the expanded poly-Q tract in the protein called huntingtin.28 Several other diseases involve different proteins with poly-Q tracts but exhibit a similar mechanism of pathology. It was also demonstrated that some SSRs not linked to the specific disease are toxic to cells when overexpressed and/or lead to protein aggregation.29–31
These and the other facts indicate that accumulation of the amyloid-like aggregates is a pathological process. This notion is further confirmed by the existence of mechanisms preventing amyloid-like protein aggregation, such as a specific chaperone preventing aggregation of excess α-globin chains.32 As misfolded and potentially aggregating proteins are usually accumulated during aging, it is an intriguing possibility that aging could promote prion-like pathologies. Indeed, some aggregation-related diseases (e.g., Alzheimer's disease) in humans are frequently associated with advanced age.
Exact mechanism of cell death in amyloid and neural inclusion disorders remains unknown. At least in case of mammalian PrP, it is certainly not due to lack of the normal protein function, as deletion of the PrP-coding gene does not cause a disease in mice.33 For Huntington's disease, it is proposed that aggregates sequester some essential cellular proteins.34–37 Poly-Q constructs introduced into Caenorhabditis elegans induce heat shock response at the stringency proportional to the length of the poly-Q stretch,38 and disrupt the global quality control of protein folding, possibly by interfering with the disposal of misfolded proteins.34 There is evidence that PrP and some other amyloidogenic proteins trigger cell death via apoptosis or autophagy.39–42
Aggregated (prion) forms of the yeast proteins Sup35 and Ure2, called respectively [PSI+] and [URE3], are not found in the natural, industrial and clinical isolates of Saccharomyces yeast,22,43,44 consistent with the possibility of their pathogenicity. However the prion form of Rnq1 protein, called [PIN+], was found in a few natural isolates.22,44 [URE3] decreases the growth rate of yeast.22 While [PSI+] does not affect growth rates of exponential cells,45 some [PSI+] strains exhibit facilitated cell death in the deep stationary phase, similar to apoptotic processes in higher eukaryotes (Y. Chernoff, J. Kumar, and G. Newnam, unpublished data). Activation of the apoptosis-like programmed cell death pathways in the starving yeast cells has been reported previously (refs. 46–48). Some combinations of [PSI+] and [URE3] isolates exhibit the synthetic lethal or sublethal interactions.49
Overproduced Sup35 protein or fragments containing the Sup35 prion domain (Sup35N) are toxic to the [PSI+] cells, or (at very high levels) to the [psi−] cells containing the [PIN+] prion, that facilitates de novo [PSI+] induction.17,50–52 This toxicity is not simply due to accumulation of excess protein per se, as it is controlled by the same protein regions that are involved in prion formation, and is not seen in the [psi− pin−] background.17,52 It is shown that accumulation of aggregated Sup35 in the prion-containing cells is associated with cell death.53,54 This somewhat parallels mammalian prion diseases, where PrPSc-related pathology is usually detected only in neurons, cells known to produce mammalian prion protein (PrP) at high levels.3
Some mammalian amyloidogenic proteins are also toxic to yeast. The poly-Q expanded fragment of human huntingtin, fused to the green fluorescent protein (GFP) generates aggregates and causes toxicity only in yeast cells containing the endogeneous QN-rich prions, [PIN+]55 or [PSI+],56 which manifest themselves as susceptibility factors for a poly-Q disorder. Prion-dependent poly-Q cytotoxicity in yeast is associated with a defect of endocytosis, apparently due to sequestration of some actin-assembly proteins, involved in formation of the endocytic vesicles, by poly-Q aggregates.57 Sup35 aggregates also interact with some cytoskeletal proteins involved in the endocytic/vacuolar pathway, and cytotoxicity of overproduced Sup35 is increased in the strains with the cytoskeletal defects.53,58 Expression of mammalian α-synuclein in yeast leads to its aggregation and cytotoxicity with some characteristics of apoptosis.59 Taken together, these data confirm that accumulation of prions and other amyloidogenic protein in yeast may lead to the pathological consequences, and establish yeast prions as appropriate models for studying the mechanisms of amyloid cytotoxicity.
One argument in favor of the adaptive role of prions is evolutionary conservation of prion-forming properties of some proteins. Prion properties of Sup35 are conserved in various species of Saccharomyces (see ref. 59a), as well as in Candida albicans and Pichia methanolica, budding yeast species that are distantly related to Saccharomyces cerevisiae.43,60–63 Comparison of the Sup35 sequences among the different isolates of S. cerevisiae and between the sister species of S. cerevisiae and S. paradoxus demonstrates that while the prion domain (Sup35N) is evolving much faster than the C-proximal release factor domain (Sup35C), sequence of Sup35N still remains under the purifying selection pressure, confirming that this region of the protein is playing a certain positive biological role.64 As the ability to form a prion is the only function of Sup35N known thus far, the simplest logical explanation would be that the ability to form a prion is adaptive under certain circumstances. Remarkably, the highest level of sequence conservation was observed within two subregions of Sup35N, the N-proximal QN-rich stretch (QN) and the region of oligopeptide repeats (ORs, see Fig. 1), which are still clearly seen in the distantly related budding yeast species of Candida and Pichia, despite low overall conservation of the Sup35N aa sequence (reviewed in ref. 65). Both subregions play a major role in prion-related properties of Sup35N (reviewed in ref. 7). However, these observations can argue in both ways, as repetitive structure of OR region per se is not a requirement for prion propagation.66 Then, conservation of OR region (and possibly of QN) could be related to some unknown function of this part of the protein that is distinct from its prion-propagating ability.
The Sup35N region of the distant relative of budding yeast, the fission yeast Schizosaccharomyces pombe, does not contain QN and ORs (Fig. 2) and exhibits essentially no aa identity (only 18%) with the corresponding domain of S. cerevisiae, while Sup35C remains highly conserved (64% identity).65 Likewise, neither sequence nor aa composition patterns of Sup35N are conserved between yeast and mammals, and the capability of Sup35 homologs (usually called eRF3) from species other than budding yeast to form prions is yet to be proven (Fig. 2). However, while aa composition of the Sup35N regions of higher eukaryotes is different from yeast Sup35N, it is still highly unusual. For example, N-terminal domain of the Sup35 homolog from mouse and human (GSPT1) contain a high percentage of P, S and G residues (10%, 15% and 20%, respectively). Instead of the QN and OR, mammalian eRF3 proteins contain poly-G and/or poly-S tracts. In mammals with two different eRF3-coding genes, all GSPT1 orthologs contain both poly-G and poly-S, while GSPT2 orthologs contain only poly-S. These homopeptide regions are usually coded almost exclusively by identical repeated trinucleotides, suggesting that they originate from trinucleotide expansions. Recent data confirm that the poly-G expansion can indeed occur in GSPT1 and is associated with susceptibility to gastric cancer.67 Obviously eRF3 homologs of higher eukaryotes possess some unusual properties, although it remains to be seen whether these properties involve an ability to form amyloids.
At the current level of knowledge, it can not be ruled out that conservation of the Sup35N aa composition in budding yeast or unusual features of the aa composition of this region in other organisms are associated with its unknown function that is not directly related to prion formation. A variety of cellular proteins interact with Sup35N and/or Sup35M regions.5 It is possible that Sup35N influences a function of the whole protein or targets it to a specific cell compartment. Indeed, the deletion of Sup35NM coding region leads to an alteration of the sexual cycle in Podospora,68 implying that this region is not completely irrelevant to the cellular function of the protein.
The first example of a prion having an adaptive biological function is [Het-s] of Podospora that controls vegetative incompatibility.69 A cytoplasmic contact between the prion-containing and prion-free mycelia results in degeneration of the latter one. In this way, [Het-s] controls vegetative incompatibility, an adaptive trait in Podospora. Moreover, after meiotic division [Het-s] prion kills spores containing a het-S allele that is incapable of producing the prion state.70 [Het-s] is abundant in natural Podospora populations. As adaptive function of [Het-s] is achieved via cytotoxic effect, [Het-s] combines features of both “prion pathology” and “adaptive prionization” models.
Role of [Het-s] in cytoplasmic incompatibility is related to one general characteristic feature of amyloids, that is, to a high level of sequence-specificity in amyloid propagation. While proteins of different sequences may possess amyloid properties, only molecules that contain the amyloid-forming domains of nearly identical sequences can join any given amyloid fiber. Recent data show that at least some amyloids are assembled together via parallel β-sheets, for which identity of aa sequences involved in β-sheet formation is extremely important.11,71 In terms of their stringency, sequence identity requirements for amyloid formation are not dissimilar from the rules that govern complementarity of DNA strands. These requirements may explain so-called “species barrier” in prion transmission, preventing transmission of the prion state between the divergent prion domains (reviewed in ref. 65). Sequence-specificity makes prions a useful tool for the self/nonself recognition systems, as demonstrated by the example of cytoplasmic incompatibility in Podospora.
In higher eukaryotes, the stress such as heat shock is followed by formation of the nuclear and/or cytoplasmic stress granules (SG).72 Cytoplasmic SGs contain transcripts associated with 40S ribosomal subunits (48S complexes), unable to initiate translation in stress conditions. SG assembly is mediated by the RNA-binding protein TIA-1,73 which contains the C-terminal RNA recognition motif and Q-rich domain (Fig. 3A) similar to prion domains of yeast prion proteins. Deletion of Q-rich domain blocks SG formation after arsenite-induced stress in the mammalian cell culture whereas substitution of TIA-1 “prion” domain for Sup35 prion domain (PrD) restores SG production. However in contrast to prion formation, TIA-1 aggregation and SG assembly are reversible after return to normal conditions72 (Fig. 3B). Therefore, SGs provide an example of labile and economical post-trancriptional regulatory and protective mechanism contributing to the cellular function in stress conditions and based on prion-like properties.
There are several other examples of the protective mechanisms based on amyloid properties. Embryos of the fish Austrofundulus limnaeus are surrounded by an egg envelope composed of two proteins that together form a structure similar to amyloid fibrils.74 Another fish protein, type I antifreeze protein that is normally a-helical, is converted into an amyloid upon freezing, that may possibly play a protective role by inhibiting ice formation.75
As aggregation of the yeast prion proteins is increased in the stationary or non-dividing cells,54,76,77 one attractive speculation is that reversible PrD-mediated aggregation is used to protect some important proteins (e.g., Sup35) during unfavorable conditions.
Ability of prions to fix and “memorize” protein conformational changes make them ideal candidates for the role of memory molecules. Indeed, it has been hypothesized that a prion-like domain of the neuron-specific isoform of cytoplasmic polyadenylation element binding protein (CPEB) is connected to long-term memory in the shellfish Aplysia.78
There is a number of other polymerized proteins that exhibit similarities to amyloid fibers, for example the spider silk protein, spidroin,79 whose adaptive role in spiders is evident.80 It has recently been shown that one of the mammalian proteins involved in melanin production adopts an amyloid structure, so that amyloid polymers likely serve as a scaffold for melanin polymerization81 (Fig. 4). This is a first clear evidence for the positive biological role of amyloids in mammals, although it is not known whether this specific kind of amyloid possesses prion properties.
There also are examples of a beneficial role of amyloid-like aggregates in bacteria, such as facilitation of biofilm formation in E. coli by the extracellular self-assembly of the major curli protein, CsgA, containing PrP-like oligopeptide repeats,82 into typical amyloid fibrils.83 Amyloid-forming proteins of Streptomyces coelicolor, called chaplins, are essential for aerial growth.84 Moreover, it has been hypothesized85 that amyloid-like formations played an important role in the emergence of the primordial membranes and other structures at the early steps of the biological compartmentalization (reviewed in ref. 7).
Numerous attempts to identify an adaptive function of the prion state were made in case of Sup35 (eRF3), which is a translation termination factor. Formation of [PSI+] prion decreases supply of functional Sup35, leading to efficient read-through of the nonsense-mutations within ORFs. It remains unclear to which extent termination at the normal terminators, usually protected by nucleotide context,86 is affected by [PSI+]. In some genotypic backgrounds, presence of [PSI+] induces heat shock response87 and increases resistance of yeast cells to some stresses.88 Although “protective” in the artificially generated laboratory situations, such abnormalities in Hsp levels would not likely be adaptive in the long run in nature.
Systematic comparison of a variety of phenotypes (such as resistance to certain toxicants, etc.) between several isogenic pairs of [PSI+] and [psi−] strains has shown that the presence of [PSI+] was beneficial in some conditions for certain genotypes.23 However, ancestors of these laboratory strains went through multiple rounds of mutagenesis and could therefore contain unidentified nonsense-alleles. While suppression of such alleles could be beneficial for these specific strains in the laboratory, the question remains whether or not this is directly applicable to natural conditions.
It was proposed23 that the presence of [PSI+] could increase the “evolvability” of the yeast population and facilitate adaptation to environmental changes by generating new protein products from ORFs containing nonsense-mutations, weak terminators or frameshifting-prone sequences. Such a mechanism could in principle be applied to activation of the silent pseudogenes.89,90 As an extension of the modular principle in molecular evolution,91 one could suggest that new genes can be created through recombination of inactivated (pseudogene) copies, which often have no introns and are “locked” by nonsense and frameshift mutations. As pseudogenes are not functional, they can easily accumulate new mutations potentially generating new functions.92 Sporadic activation of pseudogenes through nonsense or frameshift suppression allows natural selection to choose combinations of mutations having beneficial effects. Analysis of whole genomes has revealed a number of cases, which can serve as examples of possible pseudogene resurrection.93
If [PSI+] decreases termination efficiency and therefore allow pseudogene expression, such read-through events may take place at a frequency of at least one per every million years, as suggested by the quantitative model.94 However, mutations in the genes coding for the components of translation machinery may have the same effect.5 It is therefore not clear whether the proposed mechanism is specific to a prion. Although mutated translational components are likely to turn detrimental in natural environments, so is [PSI+], judging from analysis of the natural yeast isolates.22,43,44
One potential advantage of [PSI+], not shared by most of the abovementioned gene mutants, could be that it is a dominant omnipotent suppressor affecting both termination and frameshifting. Another possibility that would give [PSI+] a preference at the population level over other mechanisms causing nonsense readthrough could be an easier transition between [psi−] and [PSI+] states. However, frequencies of spontaneous acquisition and loss of the typical “strong” [PSI+] variants are quite low, making them unlikely candidates for such a role. It is therefore possible that increased adaptability could be associated not with the “conventional” stable prion variants used in most laboratory experiments, but with the variants maintained only in certain conditions and eliminated after conditions are changed. Proof of the existence of such conditionally stable [PSI+] variants has been provided recently by identification of the [PSI+] isolate that can be maintained only at high levels of the chaperone Hsp104.95 It still remains to be shown which (if any) conditions in nature could favor maintenance of such transient variants of [PSI+].
While strong experimental data support “prion pathology” model, evidence in favor of the “adaptive prionization” model is of rather circumstantional nature. Most examples of the proven biologically positive effects of amyloid-like formations (melanin biosynthesis, stress granules, etc.) are so far dealing with the nonprion aggregates. The only prion that is clearly documented to play a biologically positive role in natural conditions, [Het-s] of Podospora, ironically does so by killing a nonprion partner.
However, one should remember that the majority of the known prions were identified by chance, due to extreme phenotypic effects caused by the corresponding proteins in the prion form, such as fatal transmissible disease in case of mammalian PrP or translation termination defect in case of yeast Sup35. It is possible that we are so far dealing only with a very top of the iceberg, and a large number of prion-like phenomena are still waiting for their discoverers. If prions are to be considered as “mutants” occurring at the protein level,6 one should not expect that randomly chosen mutations would frequently turn beneficial for the organism. Rather, the majority of them would be expected to have either deleterious effect or no effect, as in case of DNA mutations. However, it does not exclude a possibility of some beneficial changes occurring by this mechanism that could be identified in the future.
Another possible dimension of this story is that beneficial effects could be associated with the transient prion variants, as hypothesized above in case of [PSI+], while the stably propagating and usually toxic prions might represent by-products of these processes. Normal cellular functions of Sup35 and Ure2 are decreased in the prion state, suggesting that transient formation of the prion-like multimers may serve as a mechanism of feedback regulation. This notion is supported by the existence of the shortened form of Ure2, generated by alternative translational initiation and lacking the prion domain.96 Likewise, existence of the shortened transcript of the SUP35 gene in certain conditions has been reported97 but never studied carefully. Many proteins involved in DNA replication, repair and transcription contain PrD-like QN-rich domains.98 In case of the yeast transcriptional repressor Gal11, existence of two alternative transcripts has been demonstrated, of which the shorter one is missing two QN-rich domains and codes for the protein that manifests itself as a transcriptional activator rather than repressor.99 These data suggest that prion-like mechanisms of feedback regulation could be widespread, and this may explain evolutionary conservation of prion properties.
One should note that transient prion variants maintained only in certain conditions are hard to distinguish from both feedback regulatory circuits and so-called “heritable” modifications persisting for a few generations. Therefore, role of the transient prion variants in adaptive evolution, as hypothesized above, would be in agreement with the more general hypothesis of V. Kirpichnikov100 regarding the role of modifications in evolution. Moreover, prion model may provide a tool for even more direct relationship between phenotypic and “genotypic” (in traditional sense) inheritance. As prion state of a protein may influence probability of prionization of another protein,17–19 this opens a possibility for concerted modification (prionization) of several proteins at once. Such a prionization network, in turn, may potentially influence a DNA metabolism and rate of “classic” mutations, in case if some of the prionized proteins are involved in DNA replication/repair. This provides a mechanism for the possible effects of the heritable protein variations on the DNA material.
We thank R.B. Wickner and G.P. Newnam for critical reading of the manuscript and helpful suggestions. This work was supported by grants ST-012 from CRDF, RAS Presidium Program “Biosphere origin and evolution” and (Lot 2006-12.2/001) from Federal Agency of Science and Innovations (to Sergey G. Inge-Vechtomov and Galina A. Zhouravleva), by grant 07-04-00605 from the Russian Foundation for Basic Research (to Galina A. Zhouravleva), and by grant R01GM58763 from NIH (to Yury O. Chernoff).
This is a modified version of the previously published manuscript: Inge-Vechtomov SG, Zhouravleva GA, Chernoff YO. Biological Roles of Prion Domains. In: Chernoff Y, editor. Protein-Based Inheritance. Austin and New York: Landes Bioscience and Kluwer Academic Press; 2007. pp. 93–105.
Previously published online as a Prion E-publication http://www.landesbioscience.com/journals/prion/article/5059