Protein-based inheritance: Epigenetics beyond 1 the chromosome

Abstract

Epigenetics refers to changes in phenotype that are not rooted in DNA sequence. This phenomenon has largely been studied in the context of chromatin modification. Yet many epigenetic traits are instead linked to self-perpetuating changes in the individual or collective activity of proteins. Most such proteins are prions (e.g. [PSI⁺], [URE3], [SWI+], [MOT3⁺], [MPH1⁺], [LSB⁺], and [GAR⁺]), which have the capacity to adopt at least one conformation that self-templates over long biological timescales. This allows them to serve as protein-based epigenetic elements that are readily broadcast through mitosis and meiosis. In some circumstances self-templating can fuel disease, but it also permits access to multiple activity states from the same polypeptide, and transmission of that information across generations. Ensuing phenotypic changes allow genetically identical cells to express diverse and frequently adaptive phenotypes. Although long thought to be rare, protein-based epigenetic inheritance has now been uncovered in all domains of life.

Our modern view of epigenetics traces its origins to Waddington, who derived the term from Aristotle’s idea of epigenesis to describe the divergence between genotype and phenotype during development (Waddington, 1942). Since then, the meaning of epigenetics has been refined to reflect growing molecular insights. It now more commonly refers to heritable alterations in gene expression that do not arise from altered DNA sequence, usually changes in chromatin such as DNA methylation, histone modification, and nucleosome positioning (Holliday and Pugh, 1975; Nanney, 1958; Riggs, 1975). One fascinating but often overlooked mode of epigenetic inheritance is not linked to changes in chromosomes. Rather, new biological traits can emerge from heritable changes in the individual or collective activity of proteins known as prions. Prions have the capacity to adopt multiple conformations, at least one of which can self-template over long biological timescales (Alberti et al., 2009; Brown and Lindquist, 2009; Chernova et al., 2017; Coustou et al, 1997; Derkatch et al., 1997; Du et al., 2015; Holmes et al., 2013; Patel et al., 2009; Patino et al., 1996; Prusiner, 1982; Volkov et al., 2002; Wickner, 1994). This unusual folding landscape allows them to serve as protein-based epigenetic elements (reviewed in Byers et al., 2014; Garcia et al., 2014; Halfmann et al., 2010; Liebman and Chernoff, 2012; Wickner et al., 2007). Efficient conversion of the native protein into the self-templating fold leads to dominance in genetic crosses and, because prions are not physically linked to chromosomes, robust transmission of their traits to all mitotic and meiotic progeny through cytoplasmic inheritance (Figure 1).

Figure 1. Transgenerational Stability of Different Modes of Epigenetic Inheritance.

Conversion between states is indicated by color changes, with the ratio of different colored populations approximating the relative stability of each mode of inheritance. Approximate mitotic stability in terms of number of generations for which a particular mode of inheritance persists are noted above arrows spanning environmental input and mitosis. Meiotic inheritance is indicated by the color of progeny in the final column and the accompanying text.

Many proteins that can act as epigenetic elements, including prions, are intrinsically disordered; that is, they do not adopt a single structure in the cellular milieu (reviewed in Wu and Fuxreiter, 2016). Despite their prevalence in eukaryotic proteomes (e.g. ~30–40% of human proteins; reviewed in Vincent and Schnell, 2016), intrinsically disordered sequences have typically been omitted from biochemical studies for convenience (reviewed in Oldfield and Dunker, 2014). It is now clear that these sequences—long thought to be unimportant—can drive diverse biological functions (see Box 1).

Box 1. Intrinsically Disordered Domains and Epigenetics.

The conformational plasticity of a prion-like protein often stems from intrinsically disordered regions (IDRs). An emerging body of work suggested that some IDRs can give rise to liquid-liquid de-mixing (phase separation) to form membrane-less organelles within a cell. First observed over a century ago, our physical understanding of these compartments has expanded greatly with recent studies of P-granule formation in Caenorhadbitis elegans germline establishment (Brangwynne et al., 2009), among others. IDRs have also been shown to fuel smaller-scale separations that may influence information processing. Differential splicing of IDRs is ubiquitous in tissue-specific transcription (Colak et al., 2013), suggesting at least one possible means of coordinating this behavior in time and space.

These observations have given rise to a model whereby gene regulation and information flow might be driven by structural transitions involving phase transitions of nucleic acid associated proteins (Hnisz et al., 2017). In this context, topologically associated domains and nuclear compartments, which have been proposed to explain chromosome architecture and higher-order gene regulation, might be established and maintained at least in part by phase separation. Although this model largely awaits testing, it has recently been reported that the formation of condensed heterochromatin domains can be accomplished through phase separation of heterochromatin protein 1 (HP1) (Larson et al., 2017; Strom et al., 2017).

Originally seen as a fascinating but rare biological oddity, prions (Alberti et al., 2009; Brown and Lindquist, 2009; Chernova et al., 2017; Coustou et al, 1997; Derkatch et al., 1997; 2001; Du et al., 2015; Holmes et al., 2013; Patel et al., 2009; Patino et al., 1996; Prusiner, 1982; Volkov et al., 2002; Wickner, 1994; Yuan and Hochschild, 2017) and prion-like proteins (Cai et al., 2014; Si et al., 2003) have now been discovered in organisms from bacteria to humankind. Some such proteins conform to classical definitions (infectious and heritable), whereas other prion-like proteins undergo more limited conformational conversion within a cell. Many are transcription factors and RNA binding proteins that serve key roles in regulating information flow. Consequently, acquisition and loss of prion conformers creates diverse new traits that are heritable without any change to nucleic acid sequence. Here we review the biochemical and mechanistic principles underlying this form of epigenetic inheritance, its biological implications, and relationship to other forms of chromatin-based epigenetics. We also highlight recent discoveries of a broader array of ‘prion-like’ behaviors in stress responses, evolution, and gene regulation.

From Paramecia to Prions

Structural inheritance, or the transmission of an epigenetic trait based on self-templating, can be observed throughout biology. For example, preexisting surface structures in Paramecia and Tetrahymena transfer their template to future generations (Beisson and Sonneborn, 1965). At the molecular level, self-templating often occurs in the context of proteinaceous and infectious particles—or prions. This concept was first envisioned to explain the baffling transmission patterns underlying several devastating spongiform encephalopathies (e.g. Scrapie, Kuru, Creutzfeldt-Jakob). More than a century of investigation of these maladies (recently reviewed in Zabel and Reid, 2015) culminated in the discovery that a misfolded conformation of the protein PrP was responsible for their spread (Prusiner, 1982). Subsequent structural analyses revealed that the infectious and protease-resistant conformation of this protein (known as PrP^Sc) was an amyloid (Prusiner et al., 1983).

The prion hypothesis also caught the attention of yeast geneticists. Non-Mendelian inheritance of two enigmatic traits in fungi – [PSI⁺] (Cox, 1965) and [URE3] (Lacroute, 1971) – was found to arise from protein-based inheritance (Chernoff et al., 1993; Doel et al., 1994; Patino et al., 1996; Wickner, 1994). Unlike traits encoded in DNA, which are transmitted to half of all meiotic progeny, [PSI⁺] and [URE3] are inherited by all meiotic progeny. These traits could also be permanently eliminated by transient growth on medium containing low concentrations of guanidine hydrochloride (Tuite et al., 1981). Transient inhibition of chaperone activity – in particular the hexameric disaggregase Hsp104 – also led to loss of these traits (Chernoff et al., 1995) and indeed guanidine hydrochloride was later identified as an inhibitor of Hsp104 function. This sensitivity arises because Hsp104 catalyzes the fragmentation of mature fibrils into smaller infectious ‘seeds.’ These seeds are transmitted to daughter cells, where they serve as ‘replicons’ for future rounds of templating (Glover et al., 1997; Paushkin et al., 1997, Shorter and Lindquist, 2004). Prion inheritance is thus strongly dependent upon chaperone activity. For example, either excess or insufficient levels of Hsp104 can eliminate [PSI⁺] (Chernoff et al., 1995; Shorter and Lindquist 2004).

In [PSI⁺] cells the translation termination factor Sup35 switches into a prion conformer that can template its fold onto other natively folded molecules of the same protein. This self-perpetuating aggregation inactivates Sup35, promoting stop codon read-through (Liebman and Sherman, 1979)—the [PSI⁺] phenotype (Figure 2). It has been hypothesized that, as a result, proteomic complexity increases via expression of formerly silent information within 3’ untranslated regions, creating phenotypic variation from the same underlying base pair sequence (Baudin-Bailieu et al., 2014; Eaglestone et al., 1999; True et al., 2004). Transformation of naïve cells with assembled amyloid made from recombinant Sup35 causes them to become [PSI⁺], providing gold-standard proof of the ‘protein-only’ prion hypothesis (King and Diaz-Avalos, 2004; Tanaka et al., 2004). Similar protein-only transformations have now been demonstrated for several other prions, including PrP (Alberti et al., 2009; Chakrabortee et al., 2016a; Patel and Liebman, 2007).

Figure 2. Protein-based Epigenetic Inheritance Impacts Diverse Aspects of Biology.

Each panel illustrates a different example of protein-based mechanisms of inheritance. [PSI⁺]: the yeast translation termination factor Sup35 normally ensures faithful translation termination by the ribosome. However, when Sup35 is sequestered into an oligomeric prion state, the fidelity of termination is reduced. This leads to the inclusion of variants in 3’ UTRs (denoted by star). Consequently, phenotypic diversity of the population is increased. [GAR⁺]: fermentative growth of yeast produces ethanol that helps to kill bacterial competitors (noted by X’s). However, certain bacteria secrete small molecules (e.g. lactic acid) that can lead to induction of the [GAR⁺] prion. This heritably shifts the metabolism of yeast, reducing ethanol output and increasing capacity for growth on complex carbohydrates. CPEB: upon serotonin signaling CPEB oligomerizes with RNA in a self-templating complex, leading to long-term memory facilitation.

Conformational Conversion as a Mode of Inheritance

The unusual folding landscape of prion proteins allows them to transfer information across generations through the inheritance of stable and infectious protein conformations. Below we discuss the critical mechanistic features that underpin this mode of inheritance. Prions often contain biased Q/N-rich amino acid sequences that are organized in domains. These regions can drive the assembly of multiple self-templating polymorphs—so-called prion ‘strains.’ Remarkably, even subtle sequence divergence within prion domains can give rise to robust barriers to transmission between species. Most prion states replicate with the help of chaperones, providing a direct link between their inheritance and environmental stress.

Unlike globular proteins that are folded in their native states, many prion proteins contain disordered regions that are natively unfolded, but have the capacity to become hyper-structured under certain conditions. Biophysical characterization of prion assembly reveals at least three distinct states: a native fold, oligomeric intermediates, and amyloid fibers. For example, Sup35’s prion domain is initially unstructured in solution (Serio et al., 2000). Upon conversion to an oligomeric intermediate through conformational rearrangement, an amyloid fiber rapidly assembles (Glover et al., 1997; Serio et al., 2000). Fiber formation is driven by the Q/N-rich N-terminal domain of Sup35 (Derkatch et al., 1996). Indeed, the capacity of many prion proteins to self-template is driven by Q/N-rich domains that form tightly packed cross β-sheets in the assembled state (Nelson et al., 2005).

One of the great surprises in prion biology has been the existence of prion ‘strains’: distinct transmissible epigenetic activity states that are formed by the same protein. Scrapie strains with varying incubation times and pathologies were observed early on, and their stability was dependent on origin and passage conditions (Bruce and Dickinson, 1987). Likewise, different stable [PSI⁺] variants were also observed upon prion induction (Derkatch et al., 1996). Some argued that this was evidence for the involvement of nucleic acid, but in vitro assembly of purified protein led to the realization that conformational diversity gives rise to strain behavior (Kocisko et al., 1994). It is now clear that the capacity to acquire multiple, stable self-templating conformations – often associated with different phenotypes – is a shared property of many prion proteins. For example, although the core amyloid structure consisting of Sup35NM β-sheets remains relatively constant, variation in amyloid fiber length and the nature of intermolecular interfaces between oligomeric intermediates gives rise to distinct self-replication potential and phenotypes in vivo (Derkatch et al., 1996; Tanaka et al., 2004).

Even subtle variation in the sequence of a prion domain can establish a strong barrier to self-templating between orthologous proteins from closely related species. Both mammalian PrP and fungal prions exhibit such ‘species barriers.’ Expression of a Syrian hamster (SHa) PrP transgene in mice renders the animals highly susceptible to SHa prions, but does not induce conformational conversion of mouse PrP (Scott et al., 1989). In yeast, ongoing templating of Sup35 from one species does not lead to cross-templating of Sup35 from another. This barrier can be isolated to specific epitopes within the protein sequence (Chien et al., 2003; Santoso et al., 2000), highlighting the importance of sequence and conformational specificity for efficient prion propagation.

Many yeast prions characterized to date share Sup35’s strong bias towards low-complexity Q/N rich sequences (Alberti et al., 2009). However, a proteome-wide screen recently revealed that many prion-like elements do not share this amino acid composition (Chakrabortee et al., 2016a). Moreover, many do not form amyloid and some require chaperones other than Hsp104, such as Hsp70 and Hsp90, to propagate (Chakrabortee et al., 2016a). Although it remains to be seen whether these newly catalogued prions share the other biophysical properties of canonical prions, they underscore the diversity and robustness of conformational conversion as a means to spark epigenetic diversification.

Stochastic and Concerted Induction

Prion states can be acquired through either stochastic or concerted mechanisms. The former encompasses both spontaneous acquisition—by chance appearance of the prion conformer and its subsequent encounter with natively folded proteins—as well as modestly increased rates of acquisition driven by general changes in the cell’s proteostasis environment, usually due to stress or age. Concerted acquisition, in contrast, refers to active induction of the state, which can be precipitated by signaling events. These two modes are not mutually exclusive, and in fact frequently co-occur.

Stochastic conversion—often referred to as spontaneous conversion—relies on the sampling and adoption of a self-templating conformations (reviewed in Halfmann et al., 2010) in accordance with the energetics of a protein’s conformational ensemble. Once a self-templating conformer arises, it can efficiently propagate its fold onto other proteins of the same sequence within the originating cell—and occasionally spread between cells. Spontaneous conversion rates have been measured for a number of prions. For example, [PSI⁺] occurs at a rate of ~6 × 10⁻⁷ (Allen et al., 2007; Chernoff et al., 1999; Lancaster et al., 2010), [GAR⁺] at ~5 × 10⁻⁴ (Brown and Lindquist, 2009), [MOT3⁺] at ~10⁻⁴ (Holmes et al., 2013), and [MPH1⁺] at ~4 × 10⁻⁷ (Chakrabortee et al., 2016a). The likelihood of acquiring some prions can be enhanced by the presence of others. For example, cells harboring [PIN⁺] acquire [PSI⁺] at higher frequencies than [pin⁻] cells (Allen et al., 2007; Chernoff et al., 1999; Zhou, et al., 2001). Induction can also be precipitated by endogenous processes such as aging, during which rates of [PSI⁺] induction increase due to decline in autophagic capacity (Speldewinde and Grant, 2017). Indeed, interactions between prions can even give rise to new heritable traits (Nizhnikov et al., 2016). Rates of prion loss have been less well-studied, but they are often similar to the rates of acquisition (Charkrabortee et al., 2016a; Holmes et al., 2013).

Epigenetic switching between native and prion states could in principle afford increased flexibility for adaptation compared to genetic changes, which cannot readily revert if environmental changes render them no longer advantageous. Indeed, rates of both prion acquisition and loss almost always significantly exceed the rate of genetic mutation (~10⁻¹⁰ per base pair per generation; Lang and Murray, 2008). In sufficiently large populations, these relatively high rates of prion conversion mean that the population will always contain some individuals that harbor the prion state. Such ‘bet hedging’ can provide population-scale fitness advantages in variable environments (Cohen, 1966). If the prion-based traits are detrimental, then only a few individuals will be lost. However, if they are advantageous, the adaptive phenotype conferred by the prion will allow the population to survive when it would otherwise have perished. Prions are also lost at frequencies higher than mutation, providing a complementary bet-hedging mechanism if the environment again shifts such that the traits conferred by the prion are no longer advantageous. Mathematical models of this framework (Lancaster et al., 2010) have been applied to [PSI⁺], [MOT3⁺], and [GAR⁺] in yeast; each is predicted to afford greater population survivability in a range of fluctuating environments (Holmes et al., 2013; Jarosz et al., 2014a; True et al., 2004).

Prions can also be induced in response to specific signals and environmental stress. Induction can coincide with relatively high frequencies of spontaneous switching, perhaps reflecting complementary evolutionary strategies depending on the frequency of environmental fluctuation (Jarosz et al., 2014b). [GAR⁺] is strongly induced by cross-kingdom chemical communication with specific bacterial species (Figure 2; Garcia et al., 2016; Jarosz et al., 2014a), rates of [PSI⁺] induction can be increased under a variety of osmotic and oxidative stresses (Tyedmers et al., 2008), [MOT3⁺] can be induced under ethanol stress (Holmes et al., 2013), and [LSB⁺] is engaged in response to heat stress (Chernova et al. 2017). Notably, [MOT3⁺] and [GAR⁺] can also be efficiently eliminated under specific conditions: hypoxia and desiccation, respectively (Holmes et al., 2013; Tapia and Koshland, 2014). Other signals that might trigger prion gain, loss, and conformational conversion—particularly in the context of normal development and physiology—remain largely unknown, and pose an exciting topic for future study.

Transgenerational Stability

Although most mechanisms of epigenetic inheritance permit robust transmission of phenotypes through mitotic cell divisions, their stability varies widely. During meiosis, traits encoded by chromatin modification are generally lost or actively erased (Figure 1, reviewed in Rando and Verstrepen, 2007; Heard and Martienssen, 2014). The two most familiar modes of epigenetic inheritance, covalent histone modification and DNA methylation, are generally stable for a few dozen (Gottschling et al., 1990; Grewal and Klar, 1996) to tens of thousands (Hernday et al., 2002) of mitotic divisions, respectively. Although some recently discovered prion states have stabilities akin to those of histone modification (Chernova et al. 2017), many others are at least as mitotically stable as DNA methylation (Lancaster et al., 2010; Lund and Cox, 1981). In this regard, the heritability of prions differs substantially from other protein aggregates, which are often retained in mother cells (Liu et al., 2010; Zhou et al., 2014).

One of the oldest known features of prions is faithful transmission to meiotic progeny—distinguishing them from other forms of epigenetic inheritance (Cox, 1965; Young and Cox, 1971). This is particularly striking given that both the canonical epigenome (reviewed in Heard and Martienssen, 2014) and most aggregates (Ünal et al., 2011) are turned over during this highly regulated developmental process. Although the molecular mechanisms underlying such escape are currently unknown they provide an attractive area for future investigation.

Remodeling Information Flow

Prions initially garnered attention because they subvert the central dogma. We now know that many are involved in information decoding within Crick’s original framework: chromatin modifying enzymes, transcription factors, and RNA binding proteins. Remarkably this enrichment holds for both amyloid and non-amyloid prions (Alberti et al., 2009; Chakrabortee et al., 2016a). In fungi, examples include [PSI⁺] (see above), [MOT3⁺], and [SWI⁺]. [MOT3⁺] promotes facultative multicellularity by converting the Mot3 transcription factor (Alberti et al., 2009; Holmes et al., 2013); [SWI⁺] abolishes multicellular growth and is driven by templating of the Swi1 chromatin remodeling factor (Du et al., 2015). Even in bacteria, previously thought to lack prions, the Rho transcription terminator can adopt an infectious conformation, leading to global changes in the transcriptome (Yuan and Hochschild, 2017).

Signal transduction modules are also common among prions. In S. cerevisiae the [GAR⁺] prion rewires glucose sensing pathways, enabling cells to efficiently metabolize mixed carbohydrates (Brown and Lindquist, 2009). In Podospora anserina the [Het-s] prion serves as an activation trigger of cell death in heterokaryon incompatibility (Coustou et al., 1997). Here, the prion functions in a signal transduction process in which a NOD-like receptor (NWD2) controls the HET-S cell death effector (Riek and Saupe, 2016). Finally, the homeobox protein Luminidependens, which instills a memory of vernalization in Arabidopsis thaliana, has the capacity to undergo prion-like conformational conversion (Chakrabortee et al., 2016b). The biological breadth of these examples is considerable given our emerging understanding of prion-based inheritance relative to other epigenetic phenomena.

Other Forms of Protein-based Epigenetic Memory

Other types of protein-based epigenetic memory have also been reported. Most do not meet the classical definition of prions (proteinaceous and infectious), but they nonetheless fuel a diverse array of biological processes. Perhaps the most striking are the many neurodegenerative diseases that arise from the aggregation of specific proteins, such as Aβ in Alzheimer’s disease, and α-synuclein in Parkinson’s disease (Jucker and Walker, 2013). Although these syndromes are not generally infectious (for a rare example of iatrogenic transmission see Jaunmuktane, et al., 2015), the protein aggregation processes that drive them can be. Recent observations of prion-like spread of the Tau protein (Sanders et al., 2016), and of conformational diversity in aggregate Aβ species linked to specific patient phenotypes (Qiang et al., 2017) have heightened interest in this behavior.

Non-infectious protein conformational conversion can also impart biologically important epigenetic memories. In S. cerevisiae an RNA binding protein known as Whi3 allows cells to ‘memorize’ deceptive mating attempts (Caudron and Barral, 2013) by forming super-assemblies that prevent future mating. Formation of these assemblies requires a prion-like domain within Whi3. However, in contrast to prions, these super-assemblies are not transmitted to daughters, and are thus termed ‘mnemons.’ Whi3 assemblies also accumulate during normal cellular aging, and appear to explain the long-described sterility of old mother cells (Schlissel et al., 2017).

In human cells, pathogens and other danger signals activate RIG-I and NLRP3 to produce immune responses via MAVS and ASC (Cai et al., 2014). The signal transduction crucial to this process is mediated by conformational conversion of MAVS and ASC into a self-perpetuating state. This process involves pairing with a NOD motif, and is strikingly reminiscent of Het-s, despite the vast evolutionary distance between mammals and filamentous fungi (Riek and Saupe, 2016). Functional amyloid is also employed extensively, as with curli proteins in the formation of biofilms in diverse bacteria (Blanco et al., 2012) and the Rim4 RNA binding protein in the regulation of meiosis in S. cerevisiae (Berchowitz et al., 2015).

Most surprisingly, given the rich literature implicating prion-like proteins in the progression of neurodegenerative disease, conformational conversion of a translational repressor may be important for long-term facilitation at the neurological synapse. Originally discovered because its synthesis is required for long-term memory formation, the CPEB protein (Orb2 in flies, and CPEB3 mammals) contains a large Q-rich domain at its N-terminus (Si et al., 2003). CPEB has prion-like properties when expressed in yeast (Si et al., 2003), and the Q-rich domain of Orb2 is required for the formation of long-term memories in flies (e.g. of courtship behavior; Khan et al., 2015). This conversion transforms Orb2 from a translational repressor into a translational activator (Khan et al., 2015). Similar conformational conversion of CPEB3 coincides with long-term memory facilitation in mice (Figure 2; Fioriti et al., 2015).

On a conceptual level, prion conversion can be thought of as a bistable system with hysteresis (Figure 3). With increasing concentration (and/or environmental stimuli) prion proteins ultimately convert to a stable alternative fold that is capable of self-templating (seeding). This allows prion proteins to occupy two or more stable activity states. Chaperones fragment prion assemblies to generate more seeds – ‘replicons’ that drive further rounds of prion conversion. This provides positive feedback, ensuring that once a prion conformation is acquired it can be stably maintained despite fluctuations in protein concentration or inducing stimulus (Figure 3). The bistability inherent to prion-based biochemistry can thus be stably maintained over long biological timescales. Other bistable circuits can also give rise to heritable traits (Roberts and Wickner, 2003), although no natural examples have been discovered that are as stable as prions. The potential for prion-like conformational conversion to act as a general means for implementing bistable regulatory switches in gene control is an intriguing hypothesis (see Box 1), but this model awaits further experimental exploration (Hnisz et al., 2017).

Figure 3. Self-templating Conformational Conversion as a Form of Bistability and Hysteresis.

A, Schematic for conformational conversion of prion states and chaperone mediated production of prion seeds. Because prion conformers self-template in the presence of chaperones, they create an autonomous positive feedback loop dependent on the prion protein alone. Native monomeric protein (left) can be converted to the prion conformer by seeding (top), resulting in the formation of oligomeric prion conformers (right). Prion oligomers can be disrupted by chaperone activity, liberated seeds to template further conversion of native monomers. B, Bistability arising from autonomous positive feedback of prion conformers creates hysteresis. Prion seeding may also provide an ultrasensitive response where a non-linear increase in prion conversion follows rapidly once seeds are formed. The system exhibits hysteresis and the prion conformation is stable even in the absence or lesser extent of stimulus. In the context of transcriptional control, this could robustly insulate two states, ‘on’ and ‘off’ (see right-hand side of panel), from small changes in stimulus, while still permitting rapid interconversion in response to directed triggers.

Concluding Remarks

The capacity of protein sequences to form highly ordered aggregates is likely ancient. Primitive polypeptides in ‘primordial soup’ experiments can organize into amyloids (Greenwald et al., 2016), and some have speculated that this behavior played an important role in shaping the origin of life (Singh and Banerji, 2011). In extant species, many proteins now appear to have the capacity to undergo ‘infectious’ conformational conversion. This behavior is often driven by intrinsically disordered sequences (Chakrabortee et al., 2016a) – ‘dark matter’ of the proteome that is relatively rare in bacteria, but common in eukarya and scales with increasing organismal complexity (Ward et al., 2004; reviewed in Wu and Fuxreiter, 2016).

Long thought to be a biological curiosity, it is now clear that protein-based inheritance – epigenetics beyond the chromosome – is exploited by a wide range of organisms and in many biological circumstances (Si et al., 2003; reviewed in Byers et al., 2014; Halfmann et al., 2010; Wickner et al., 2007). Because the discovery of protein-based inheritance has been so closely tied to disease (Prusiner, 1982; Wickner et al., 2011), prions have often been seen as detrimental. However, just as viruses revealed fundamental features of nucleic acid-based inheritance, diseases associated with self-templating protein misfolding have revealed fundamental mechanisms that drive this ancient mode of epigenetic information transfer.

Despite rapid growth in our understanding of protein-based inheritance in the past two decades, many key questions remain. Some protein-based epigenetic states can be robustly triggered by specific stimuli such as [GAR⁺] (by cross-kingdom communication with bacteria; Garcia et al., 2016; Jarosz et al., 2014a) and the Whi3 mnemon (by deceptive mating attempts and aging; Caudron and Barral, 2013; Schlissel et al., 2017). Yet whether specific triggers for induction are common, and what their molecular origin might be, remains to be established for most protein-based elements of inheritance. Likewise, aside from [GAR⁺] (Tapia and Koshland, 2014) and [MOT3⁺] (Holmes et al., 2013), it is unclear whether most protein-based epigenetic states, which are extraordinarily stable compared to many chromatin marks, can be erased by specific signaling inputs or environmental stimuli. If such behavior is general, it would further point to the utility of this mechanism as a robust, but regulatable, means to instill and broadcast biological memories independent of chromosomes.

SUMMARY.

Self-templating changes in protein conformation can drive unusually robust epigenetic ‘memories’ that are heritable across generations. This biochemistry permits multiple stable activity states to arise from a single polypeptide sequence, fueling diverse biological functions.