Protein self-assembly: A new frontier in cell signaling

Abstract

Long viewed as paradigm-shifting, but rare, prions have recently been discovered in all domains of life. Protein sequences that can drive this form of self-assembly are strikingly common in eukaryotic proteomes, where they are enriched in proteins involved in information flow and signal transduction. Although prions were thought to be a consequence of random errors in protein folding, recent studies suggest that prion formation can be a controlled process initiated by defined cellular signals. Many are present in normal biological contexts yet are invisible to most technologies used to interrogate the proteome. Here we review mechanisms by which protein self-assembly can create a stable record of past stimuli, altering adaptive responses, and how prion behavior is controlled by signaling processes. We touch on the diverse implications that this has for normal biological function and regulation, ranging from drug resistance in fungi to the innate immune response in humans. Finally, we discuss the potential for prion domains in transcription factors and RNA binding proteins to orchestrate heritable gene expression changes in response to transient signals, such as during development.

INTRODUCTION

To adapt and survive, organisms digest a constant stream of information about their surroundings, deploying signaling mechanisms that transduce this information into changes in gene expression, morphology, and behavior. Interestingly, signaling proteins frequently harbor long intrinsically disordered sequences[1,2], that do not adopt a fixed structure in solution, but can often reorganize upon ligand binding or post-translational modification. Here we focus on additional property of disordered protein regions: the ability to drive heritable information transfer via prion-like self-templating.

Prion-like proteins can adopt multiple stable conformations, at least one of which can self-perpetuate over long biological timescales. This unusual folding landscape can drive a form of epigenetics that is both heritable and reversible. Prion conformations are dominant, converting all molecules of the protein within a cell, generally causing the protein to amalgamate into larger assemblies. Chaperones fragment these assemblies, generating ‘seeds’ that are broadcast to daughter cells, launching new rounds of replication and allowing the alternative conformer to persist in a lineage.

Because prions are transmitted extra-chromosomally, their patterns of inheritance defy Mendel’s laws. Their transmission also strongly depends on molecular chaperones, providing intrinsic links between environmental stress and their gain and loss. This balance of heritability, reversibility, and environmental responsiveness allows protein-based epigenetic elements to avoid some costs of adaptation via genetic mutation[3–5]. Such benefits are amplified when prions are elicited and erased by defined biological stimuli[6–8]. Amidst the growing appreciation of the diverse roles that protein disorder plays in biology, multiple properties of prion self-assembly provide appealing advantages in the context of signal processing and cellular ‘memory.’

Diverse consequences of self-assembly

Integrating protein self-templating into developmental processes or stress responses might at first glance seem redundant with normal signaling functions. Yet doing so provides multiple advantages. One is to provide a long-lived record of a past stimulus. For example, S. cerevisiae cells exposed to lactic acid produced by bacterial competitors induce the [GAR⁺] prion, allowing them, and their descendants, to engage an adaptive metabolic program even after the inducing molecule dissipates[9,10]. Likewise, yeast cells exposed to antifungal drugs acquire the [MOD⁺] prion, which enables heritable resistance to these compounds[11] (Figure 1).

Figure 1. Integration of prion-like behavior into signal transduction pathways.

A) Archetypical signaling pathway, a membrane receptor senses the presence of the signaling molecules and transmits the signal through a kinase cascade, culminating in a transcriptional response. B) Prion self-assembly can provide an enduring record of previous exposure, maintaining the response even after the signaling molecule is gone. C) Prion-based self-assembly can amplify a response, recruiting more signaling molecules and ensuring fidelity by reducing noise in the system. All figures were created with Biorender (Biorender.com)

Prion switching converts all molecules of a protein into the self-templating form, providing a mechanism to amplify the original signal, and a second advantage of integrating prion behavior into signaling pathways. The human mitochondrial antiviral signaling protein (MAVS), which forms prion-like aggregates upon activation by RIG-I-like RNA helicases interacting with viral RNA, provides a striking example. Prion-like assembly allows a subpopulation of MAVS to seed all remaining native protein into functional aggregates, activating the IRF3 transcription factor to drive a robust antiviral response. Self-assembly has also been proposed to facilitate propagation of the antiviral signal between adjacent mitochondria[12] (Figure 1).

Finally, prion-like self-assembly can create reservoirs of signaling molecules, such as in the Balbiani body in oocytes, the first asymmetric cellular structure in the development of many vertebrates[13]. In Xenopus, the prion-like Balbiani core protein Xvelo assembles to form a stable amyloid-like mesh, sequestering maternal RNAs and organelles, aiding the development of the embryo and perhaps passing healthy components to the next generation[14]. In a conceptual parallel, amyloid assemblies observed during stress in yeast protect metabolic enzymes required to survive and restart the cell cycle after the environment improves[15].

Hidden in plain sight

[PRION⁺] cells have the same genotype as naive cells. In many instances naive and [PRION⁺] cells also express the same amount of protein, just with a different fold and localization. Thus, many “omics’’ technologies cannot easily identify them: prions that are ubiquitous and control some of the most widely studied traits in biology remained undiscovered for decades[16,17]. In this respect, recently developed structural proteomic technologies, such as limited proteolysis–mass spectrometry (LiP-MS) offer a promising route forward, and has recently been used to resolve oligomeric conformations of prion-like proteins, such as alpha-synuclein, in living cells[18].

Acquisition and Loss

Spontaneous prion conversion can range from ~one in ten thousand in the case of [GAR⁺] or [MOT3⁺] to ~one in ten million in the case of [PSI⁺] or [MPH1⁺] (reviewed in [19]). These conversion rates are orders of magnitude higher than spontaneous mutation (~10⁻⁹), and in any sizable microbial population are easily achievable. Thus, a population may harbor many different natural prions encoding pre-adaptive states that help the population to survive under stress.

In addition, de novo inducers of some prions can dramatically increase their occurrence. The rate of prion switching can be responsive to carbon sources, metabolites from other organisms in the environment, cell cycle arrest, or exposure to antifungal drugs (as for [MOT3⁺], [GAR⁺], [ESI⁺], or [MOD5⁺], respectively). The inducer can have a remarkably complex influence on the conversion process. For example, different concentrations of lactic acid (the inducer) produced by bacterial competitors generate distinct [GAR⁺] prion ‘strains’ that continue to propagate with different strengths once the inducer is removed[20]. Thus, fluctuations in inducing signals can cause the same prion protein to acquire different stable conformations, perhaps tuned to the environmental need.

Amyloid is only the beginning

The striking amyloid fibers formed by PrP[21], the first prion to be discovered, indelibly shaped our inferences about what biochemistries might drive protein-based information transfer. Yet the expanding ‘prionome’ has revealed that prions have diverse sequence features, biophysical properties, and chaperone dependence[1]. In this sense, Prusiner’s original definition of prions -- proteinaceous and infectious particles -- was remarkably prescient.

Even among fiber architectures there is a great deal of conformational diversity, giving rise to a phenomenon known as prion ‘strains’ or ‘polymorphs’: Different heritable protein conformations produced by the identical protein sequence. This property can also be seen in the [SMAUG⁺] and [ESI⁺] prions, which form infectious gels rather than fibers[16,22], suggesting a type of structural ‘allelic variation’ that may play an important functional role.

The possibility that the molecular grammar of prions might extend beyond amyloid inspired us to engage in a phenotypic search for prions amongst all genes in Saccharomyces cerevisiae[1]. This screen capitalized on the law of mass action, and the extreme thermodynamic stability of prion conformations, using transient overexpression followed by outgrowth and dilution to induce acquisition of self-templating protein folds that remain once the level of protein returns to normal. This same type of experiment provided key evidence that Sup35 encodes [PSI⁺] and Ure2 encodes [URE3]. Following transient overexpression, 80 proteins induced long-lived persistent epigenetic states; 46 of these had dominant, non-Mendelian patterns of inheritance in genetic crosses and could be eliminated by transient chaperone inhibition. Remarkably, most did not result from amyloid fibers, nor were they Q/N rich. This group of proteins was enriched in intrinsically disordered regions (IDRs), opening the question of whether a broader array IDRs can act as prion domains, and remarkably these IDRs are often preserved from yeast to mammalian orthologs. The greatest functional enrichment was among proteins involved in signaling and downstream gene regulation, suggesting that information storage in non-amyloid prions may be a common attribute of signal transduction pathways.

A ‘prion’ within a cell

Prions can be distinguished from many other protein assemblies by their robust transmission in both mitotic and meiotic cell divisions. Yeast cells, for example, ensure the rejuvenation of their daughters by retaining age-associated aggregates and misfolded proteins, for example by tethering them to the endoplasmic reticulum or importing them to mitochondria[23]. In addition, the ER-bound chaperone Ydj1 acts a sink for these protein deposits and, together with the bud diffusion barrier[24], confines them to mother cells[25]. During gamete formation, age-associated factors are also sequestered in a ‘recycling bin’ arising from the nuclear membrane remodeling[26]. By contrast to prions, which can escape these mechanisms, these mechanisms constrain other self-templating aggregates to the cell in which they arise -- as a “prion within a cell.” Whi3 super-assembly provides a prime example. Upon deceptive mating attempts, Whi3 assembles, releasing the G1 cyclin Cln3 allowing the escape of mating-induced cell arrest. These ‘mnemon’ assemblies self-template to confer a memory of the non-productive encounter, but do not escape the cells that were deceived [27]. The degree to which the barrier can be controlled is a fascinating question for future study[28] In higher eukaryotes, similar assemblies take part of neuronal memory. Self-templating of CPEB/Orb2 in Drosophila transforms its activity from a translational repressor into an activator, promoting synapse facilitation and long-term memory[29,30].

Stability: Short- and long-term memory

Many ‘membraneless organelles,’ from P-bodies to transcription super clusters, are maintained by some degree of protein self-association. However, these structures are relatively labile, breaking and reforming over the course of a cell cycle. By contrast, although they can also be induced and eliminated, prions are more persistent, permitting them store information over long biological timescales. The stability of these self-assembling particles, and their transmissibility to daughter cells, are key factors in the duration of the memory. For instance, a stable and cytoplasmically transmitted prion can make a record that is shared with the lineage, whereas mnemons such as Whi3 provide a memory of past experiences restricted to individual cells. Phenotypic screens for prion-like elements have also suggested that some may persist over more intermediate periods. Understanding the breadth of timescales over which protein-based memory can operate is an important question for future study (Figure 2).

Figure 2. Prions and “prions within the cell”.

In prions the stimulus induce a change of protein conformation that can self-assemble and escape cellular mechanisms to tether proteins assemblies in mother cells. This allows the prion to propagate over generations and the cytoplasmic transmission of this function through mitosis and meiosis. In mnemons “Prions within a cell” the stimulus induces similar self-perpetuating assembly that is restricted to mother cells and cannot be transmitted to daughter cells. In this case the mnemons are stable over generation and confer memory only to the mother cells that have been exposed previously to the stimulus.

Where is the line between protein self-assembly and bona fide prions?

Important aspects of multiple signaling pathways such as Ras and TOR are also modulated by protein assemblies including the dwell time of signaling molecules[31], inactivating kinases[32], activating phosphatases[33], increasing the barrier to pathway activation[34], or releasing the activation factor of the pathway[27,35]. IDRs are often responsible for the generation of prions, liquid-liquid phase separation (LLPS), or gel-like solid aggregates, but these classifications overlap more than previously thought. For instance, the age of the assembly can alter the property from liquid to solid[36]. Even more surprisingly, the same domain of Sup35 that induces the amyloid [PSI⁺] prion also promotes reversible LLPS - and subsequently forms gels - under starvation[37]. Thus, the prion domain of Sup35 can produce multiple stable conformations, with different biophysical properties and stabilities, depending on the stimuli that cells have experienced (Figure 3).

Figure 3. Proteins harboring IDRs are highly dynamic.

A single protein can acquire multiple structures based on environment[37,66]. Different stimuli can induce multiple alternative conformation of the same IDR conferring different biophysical characteristics and multidimensional functions of the same protein. Stimuli can be a change in environmental conditions including stresses, nutrient availability, or interaction with ambient organisms. These stimuli can alter the cellular pH, induce PTMs or alter the interactions of the IDRs allowing them to gain multitudes of structures.

What biophysical features of intrinsically disordered sequences might spark this extraordinary structural and functional versatility? Functional motifs within IDRs have long remained obscure and identifying their multivalent interaction units is not straightforward. Moreover, their function or interactions can depend on multiple biophysical properties[38–41] that sometimes cannot be predicted directly from sequence. Many cellular assemblies are driven by multivalent interactions (reviewed in [42]). For LLPS, a stickers-spacers model, where the stickers are patches of domains separated by flexible linkers (spacers) or small sequence motifs within IDRs, provides some organizing principles. For FUS, arginines and tyrosines can act as stickers[43], whereas arginines are also important for the condensation of nuclear speckles[44]. For many RNA-binding proteins, aromatic residues can act as stickers and their number and distribution controls assembly[38]. Although certain rules can be applied on a class of proteins, other mysteries remain, such as how so many different protein assemblies can form, yet avoid mixing[45]. Stoichiometry and protein-RNA complexes competition may provide a partial explanation[46,47] but it is not presently possible to predict these behaviors from sequence alone, and these models remain to be applied to bona fide prion proteins.

Converting transient stimuli into durable changes in gene expression

In many cases, transcription factors (TFs) are the ultimate recipients of information transduced through signaling pathways, driving gene expression programs that are adaptive in stress or promote a developmental transition. Transcription factors themselves are among the most disordered and prion-like in eukaryotic proteomes[48,49]. Such sequences can be important for interactions with cofactors, transactivation, phase separation, and even interaction with target promoters[50]. Notably, they are heavily modified by kinases[51–54] and show extensive variability in splicing[55,56], providing multiple opportunities to regulate these emergent behaviors.

Within TFs, the minimal functional units are a DNA-binding-domain (DBD) and an activation domain (AD). ADs are usually located within IDRs. Yet their function can be separated from large swaths of the disordered sequence. Screening of TF fragments libraries using an assay known as tAD-seq[57] systematically mapped ADs. Some are enriched in poly-Q sequences, or phase separate with Mediator in transcriptional activation[58]. Yet these experiments also revealed that phase separation is often driven by parts of the IDR that that are separable from ADs. Future work will be needed to investigate whether these regions also harbor prion-like activity – as might be expected given their sequence enrichments.

In yeast, where many more prions have been characterized and studied, the answer to this question is an unambiguous yes. Indeed, transcription factors are highly enriched among prion proteins[1,59,60], and environmental signaling cues can induce several to adopt a self-templating fold. For example [MOT3⁺] can be induced by ethanol exposure and eliminated by hypoxia[61,62]. These shifts in TF conformation can transform biological phenotype, for example controlling whether cells grow as individuals, or part of a community[63].

Regulation of transcription downstream of signaling cascades is not solely driven by TFs, but also depends heavily on chromatin environment and organization. Prions such as [ESI⁺] illustrate how signaling can also impact this aspect of gene control via protein self-assembly. Upon cell cycle arrest, the [ESI⁺] prion scaffold protein Snt1 adopts a self-templating fold, evicting repressive factors to promote expression of subtelomeric genomic information that would normally remain silent. In this way, [ESI⁺] promotes the transgenerational inheritance of active chromatin, something thought to be impossible in the absence of a canonical read-write mechanism. As a result, cells and their descendants express many stress response genes, providing strong resistance against many environmental insults, including antifungal drugs used in the clinic[22].

Perspectives and Future Directions

With increasing appreciation that protein aggregates can encode novel functions, beyond driving cytotoxicity, there has been an explosion of interest in prion-like self-templating. Prions can record multiple ‘bits’ of conformational information in their different folding states while simultaneously reducing signaling noise[64]. Although such elements to date have been discovered in budding yeast, the intrinsically disordered sequences that promote prion-like behavior are even more abundant in metazoan proteomes. Our current understanding of prion function in microbes suggests a myriad of potential functions. Indeed, the few examples known in metazoans portend strong biological impact. Protein self-assembly assembly has now been implicated in T-cell receptor signaling[65], innate immunity[12], development[33], and activation of growth pathways[31]. Many more examples likely remain to be discovered. Owing to the conformational bistability inherent to prion conversion[19], this simple information storage function could extend, via trigger waves and other related phenomena, to information transfer over long distances such as in an axon or a syncytium. The surplus of prions among TFs may prove important during development, where instilling a molecular memory of past morphogen exposures would ensure the maintenance of robust developmental trajectories. Unifying our understanding the diverse forms of protein self-assembly, revealing for their functions, and developing tools to identify prion-like behavior across the tree of life promises to reveal a rich biology that has long been hidden in plain sight and now awaits discovery.