Roles of Nucleic Acids in Protein Folding, Aggregation, and Disease

Abstract

The role of nucleic acids in protein folding and aggregation is an area of continued research, with relevance to both understanding basic biological processes and disease. In this review, we provide an overview of the trajectory of research on both nucleic acids as chaperones as well as their roles in several protein misfolding diseases. We highlight key questions that remain on the biophysical and biochemical specifics of how nucleic acids have large effects on multiple proteins’ folding and aggregation behavior, and how this pertains to multiple protein misfolding diseases.

Graphical Abstract

Glossary of terms on protein folding and aggregation

To introduce the non-expert, we have compiled a definition of fundamental terms used throughout this review. Misfolded proteins are any number of intermediate states of a protein that are “non-native” conformations induced by external stressors like oxidation, heat, or the result of mutations, and are typically aggregation prone. Some protein folding intermediates are on-pathway, meaning that they are a required step to bridge the unfolded and folded states of the protein, while some are off-pathway, meaning that they are misfolded states that do not lead to productive folding. Misfolded proteins tend to accumulate with varying stoichiometries and form oligomers. While oligomers can refer to either folded or misfolded proteins, here we use it to describe soluble assemblies of non-native, misfolded proteins and cofactors (like nucleic acids) that have ≥2 misfolded protein units. The process of oligomerization refers to the mechanisms by which proteins self-assemble. Protein aggregates/aggregation can be broken into two sub-parts. The first involves the formation of large, ordered or amorphous insoluble conglomerates of misfolded protein is often referred to as general protein aggregation, amorphous aggregation, or sometimes off-pathway aggregation. Oligomerization that results in amyloid fibers – large, uniform, insoluble, and ordered protein aggregates with repeating cross β-sheet secondary structure – can be referred to as amyloid aggregation, fiber formation, or on-pathway aggregation.

Protein chaperones

In the 1950–60s, the Anfinsen model postulating all the necessary instructions on a protein’s proper fold was contained in its sequence was widely accepted. However, recreating this in vitro in intervening years proved difficult for many proteins. The emergence of protein chaperones in the 1970s and 80s began to challenge this dogma.¹ Chaperones are a class of proteins that prevent the aggregation and/or aid in the folding of client proteins². They are vital to life as there is a subset of proteins that cannot fold properly in the absence of chaperones.^3–6 Moreover, chaperones are responsible for promoting a healthy folding environment in the face of cellular stress.⁷ The lifetime of any individual protein can be heavily influenced by chaperones from its synthesis to its eventual degradation (Figure 1)².

Figure 1.

Potential life span of a generic protein.

A. Nascent protein can misfold or attain their native structure spontaneously or be assisted by ATP-dependent chaperones. Once in its native state a protein will exist for a period of time before being degraded or can potentially misfold as a result of stress. B. ATP-independent holdases can bind the early stages to prevent large aggregates from forming. Soluble oligomers and aggregates can be disassembled and refolded or shuttled to protein degradation systems. C. In aging and disease, cellular stress becomes chronic and overwhelms the chaperone system. This can lead to cell death as chaperones cannot buffer the effects of misfolded protein, resulting in irreversible aggregation.

Chaperones are often grouped into two classes, referred to as “foldases” and “holdases”. Foldases – including the Hsp60, Hsp70, and Hsp90 classes – are large molecular machines that use ATP as an energy source to drive large conformational changes⁷. These conformational changes enable the foldase to have a conformational cycle that promotes different binding and folding states of the client protein^6–8. These foldases also have the ability to prevent protein aggregation, but this second task is also performed by ATP-independent “holdases” (Figure 1).

Holdases include many disparate proteins, such as the small heat shock proteins^{9, 10}, the α-crystallins^11–13, Hsp33^14–16, and HdeA^{17, 18}, among others^19–22. Holdases are thought to act in a more mechanistically simple manner, by acting as a first line of defense to prevent widespread protein aggregation, later transferring their clients to the foldases for protein refolding. Although not requiring ATP, holdases can still be mechanistically complicated, using conditional disorder or heterogeneous oligomerization as regulatory mechanisms^{9, 23}. Some holdases do allow protein folding while bound, such as Spy^{24, 25}.

There is a third class of less-discussed chaperones that do not fit easily into known holdase or foldase classes, including the histone chaperone DAXX²⁶, and nucleic acid G-quadruplexes which will be discussed below. These nucleic acid structures are still relatively underexplored in their activities but can rescue protein folding intermediates to their native states and improve protein folding quality.

Polyanions in protein aggregation and folding

Polyanions in general appear to be highly effective at modulating protein aggregation, in both a positive and negative sense. There are numerous studies that demonstrate synthetic polyelectrolytes and polyanionic polymers can prevent aggregation of antibodies and other protein clients^27–30. Similarly, naturally occurring polyanions can also prevent protein aggregation. Polyphosphate on its own can prevent aggregation, proteins that are rescued by polyphosphate can be refolded by the Hsp70 foldase system, and polyphosphate production is induced by oxidative stress^{31, 32}.

Polyanions can also directly affect the folding and stability of proteins. The P22 Arc repressor was shown to have its folding accelerated not just by its native binding target, DNA^{33, 34}, but also (initially by) several other polyanions, including polyvinyl sulfate, heparin sulfate, and polyglutamate³⁴. Fibroblast growth factors have long been shown to have their activity and stability dependent on binding to polyanions³⁵.

However, the role of polyanions in protein folding of some proteins is not uniform and is countered by their potency in instigating protein aggregation. In a study of protein stability of four different basic proteins with multiple polyanions, it was shown that the effects of the polyanions were highly variable, depending partly on the hydrophobicity of the polyanion in question³⁶. Heparin is widely used to induce the fiber-like amyloid formation of many disease-related proteins^37–39. Polyphosphate, despite its chaperone abilities, appears to instigate amyloid aggregation to an even larger extent than heparin^40–42. Therefore, the roles of polyanions in protein folding and aggregation are complicated and the balance of their roles in the cell is still largely unknown.

The Ribosome as a Chaperone and the Emergence of Nucleic Acids as Chaperones

The ribosome is where proteins are synthesized in the cell. It is a ribozyme made up of both protein and RNA components, with the primary catalytic activity carried out by the RNA portion^43–45. It may come as little surprise that the center of protein synthesis is also a hub of chaperone activity, referred to as Protein Folding at the Ribosome (PFAR)^{46, 47}. PFAR’s importance is highlighted by defects being involved in prion propagation and other amyloid diseases.^47–50 Nascent proteins do not have the entire context (whether spatial or sequential) to properly fold and can aggregate more easily⁴⁶. Because of this, chaperones (trigger factor, Hsp70s, etc.) bind to the ribosome and associate with nascent proteins as they emerge⁵¹. The ribosome itself can also assist in nascent protein folding via electrostatic interactions in the exit tunnel⁵². The exit tunnel may also contribute to the protein folding activity by restricting promiscuous interactions with nascent protein outside the channel^{53, 54}. Although the exit channel is important to nascent protein stability, it appears the true power of PFAR does not reside there. Interestingly, the source of PFAR is concentrated in the rRNA at the ribosomal surface near the peptidyl transferase center (PTC).

The Das Gupta group was among the first to show that the ribosome itself was able to chaperone protein clients. For their initial work, they used E. coli ribosomes (70S ribosome) and denatured enzymes to determine if the presence of ribosomes could reconstitute their activity^55–57. They determined the activity of denatured enzymes was rescued by the presence of ribosomes, and that the clients were able to fold directly on the ribosome. They also determined enzyme clients could either fold on the ribosome and remain bound in an active state, as was the case of alkaline phosphatase. Enzyme clients could also be released from the ribosome in active form, which was the case for glucose 6-phosphate dehydrogenase. This was intriguing because it suggests the ribosome could bind/unbind denatured clients to release them in an active form similar to foldase chaperones⁵⁶.

Initially, it was unclear whether it was chaperone contamination (a variety of which bind to the ribosome), ribosomal proteins, or rRNA responsible for the chaperone activity. To determine which component was responsible for the chaperone activity of the ribosome, they isolated the rRNA via phenol extraction⁵⁸. They found that the rRNA component of the ribosome had similar chaperoning activity as the intact ribosome. When they denatured enzymes in a sample treated with RNase, there was no recovered enzyme activity, similar to samples containing no rRNA. To find which rRNA was responsible for PFAR, they sedimented and separated the rRNA components, and determined that only the 23S rRNA was necessary to recapitulate the chaperone activity of the intact ribosome (Figure 2)^59–61.

Figure 2.

Discovery of domain V as a protein chaperone.

It was discovered ribosomal extracts were able to refold denatured enzyme. The protein refolding activity was eventually isolated to the 50S subunit, specifically the highly conserved domain V of the 23S rRNA.

The 23S rRNA (28S in eukaryotes) comprises the highly conserved PTC⁶². The PTC can be inhibited by a variety of small molecules and mutations at this site can impart antibiotic resistance. The Das Gupta group took advantage of this specificity to narrow down the region of chaperoning activity on 23S rRNA by using erythromycin and chloramphenicol, translation inhibitors with known binding sites on the PTC⁶³. They refolded enzymes in the presence or absence of the inhibitors, and the presence of either inhibitor effectively prevented protein folding. It was known that both inhibitors also interacted with domain V of 23S rRNA at the highly conserved large loop region. Mutations made to this loop region and domain V drastically decreased or ablated protein folding activity^{64, 65}. Further experiments with primer extension analysis and cross-linking mass spectrometry confirmed domain V was the site of PFAR⁶⁷. Mass spectrometry experiments showed that amino acid residues bound by 23S were not solvent accessible in a folded state, suggesting that the RNA bound unfolded or partially folded states of the protein. The peptide fragments, from two different proteins, also bound to similar regions of domain V which are located at the surface of the 50S subunit⁶⁶.

Domain V is highly conserved across organisms, and the larger 23S rRNA/PTC subunit is considered the oldest part of the ribosome, thought to be linked to the first common ancestor of all living organisms⁶². The inherent chaperone activity of the subunit suggests it co-evolved along with the PTC ribozyme activity. Quite remarkably, this inherent chaperone activity is ATP-independent. This is in contrast to the well-studied foldases that require ATP to both properly fold their clients and/or be in a binding competent state⁶⁷.

Although 23S RNA has been the focus of chaperoning activity on the ribosome, other RNAs later were discovered to have chaperone function, such as the 5S RNA ribosomal RNA⁶⁸. As such, it remained possible that the roles of nucleic acids as chaperones could be considerably broader than domain V on the ribosome.

Non-23S Nucleic Acids as Chaperones

Although much of the early work surrounding nucleic acids and their roles in protein folding/aggregation focused on rRNA, broader evidence of this activity had been known anecdotally for decades. Some of the earliest work in the field described the ability of nucleic acids to prevent or nucleate microtubule polymerization in both a sequence and concentration dependent manner⁶⁹. Later it was shown that nucleic acids could facilitate the folding of the Arc repressor, where in the absence of nucleic acid or polyanion the dimerization reaction was at least 30x slower^{33, 34}.

Isolating recombinant proteins often involves tethering an additional highly soluble protein tag such as maltose binding protein or small ubiquitin-like modifier protein (MBP and SUMO respectively) to aid in the solubility of isolated recombinant protein⁷⁰. Highly charged solubility tags are more proficient in assisting the target protein’s solubility, with highly negatively charged tags generally being even more effective^{70, 71} In a novel approach to assisting recombinant protein expression, the Seong laboratory expressed a variety of proteins fused to an RNA binding domain to take advantage and uncover the chaperoning capabilities of RNA (Figure 3)^{72, 73}.

Figure 3.

RNA can assist in protein folding and stability.

A. Protein solubility tags are often used to enhance protein folding in protein purification. It was found that by fusing the tRNA binding domain of LysRS to non-nucleic acid binding proteins, they could enhance the yield of soluble protein compared to maltose-binding protein (MBP). If the domain cannot bind tRNA the folding enhancement does not occur. B. C5 protein was thought to be an RNA chaperone itself, facilitating the folding of M1 RNA. However, if RNA binding sites are mutated, the protein becomes unstable and aggregates. This indicates M1 RNA is integral to the solubility and stability of C5. C. In a study looking at how digesting nucleic acids out of cell lysates would affect protein stability, it was found that RNA digestion induced widespread aggregation, even in non-RNA binding proteins.

They fused the RNA binding domain of lysyl tRNA synthase (LysRS), which specifically binds tRNA^Lys to their target proteins. They found that in comparison to a control solubility tag MBP, LysRS fusion proteins were more soluble and folded more competently. To isolate that RNA-binding was responsible for the chaperoning effects, they used only the RNA binding domain of LysRS and mutated key residues to inhibit RNA binding. The presence of the tag alone did not enhance chaperone activity, nor did RNA-binding deficient mutants in the presence of RNA. This demonstrated that RNA binding was necessary for the enhanced folding and solubility properties observed with LysRS fusion proteins⁷³.

It had already been shown that RNA present in ribozymes can chaperone proteins (e.g., ribosomal domain V), but to this point, it was assumed that proteins chaperoned the RNA structures rather than the other way around. Seong group focused on an ubiquitous and highly conserved ribozyme, ribonuclease P, to determine how its M1 RNA chaperoned its evolutionary partner binding protein, called C5 (Figure 3B)⁷⁴. In vitro, they determined that the folded state was enhanced by the presence of M1 RNA and again that point mutations to either the RNA or protein to disrupt protein-RNA interactions negatively affected protein folding. To test whether these effects persisted in vivo, they used a dual expression system so that M1 RNA and C5 protein were separately inducible. They showed that when M1 RNA was not overexpressed alongside C5, the protein was more likely to aggregate. The aggregation of C5 could not be rescued by co-expressing mock plasmid vectors or by expressing tRNA^Lys, so the folding of C5 was specifically dependent on wt-M1 RNA⁷⁴. Not only did this work demonstrate that RNA was chaperoning its natural protein binding partner, but it also hints at the co-evolutionary link between an early RNA world with modern protein-based biology⁷⁵.

More studies continued to broaden the roles of nucleic acids even further. In one example, the folding and stability of the RNA-binding HIV protein, TAT, was regulated by its RNA binding partner, TAR⁷⁶. A separate study later showed that RNA is widely important for protein solubility: adding RNase to degrade the RNA in eukaryotic lysates caused massive protein aggregation⁷⁷. Intriguingly, the aggregation was not restricted to proteins with recognizable RNA-binding domains. Many proteins with no apparent RNA binding capacity also aggregated, suggesting a more widespread role for protein chaperoning than previously appreciated⁷⁷ (Figure 3B).

We then began investigating nucleic acids as chaperones, testing various DNA, RNA, and polynucleotide sequences for their ability to prevent protein aggregation. We discovered that nucleic acids, particularly polyU RNA, were more effective at preventing protein aggregation than known protein chaperones. Moreover, PolyU RNA could facilitate the transfer of denatured, bound proteins to the Hsp70 system for refolding.⁷⁸

Subsequent research by our lab utilized a thermal aggregation screen of 300+ randomized ssDNA sequences to explore if there was a sequence component to chaperone activity. Bioinformatics revealed that sequences enriched with a polyG motif exhibited strong chaperone activity. These sequences were found to form G-quadruplexes, unique secondary structures composed of stacked guanine tetrads that are stabilized by cations (typically potassium). G-quadruplexes were more effective in preventing protein aggregation compared to ssDNA and dsDNA controls, with RNA versions of the G-quadruplex sequences exhibiting similar activity.⁷⁹

Additionally, G-quadruplex-forming sequences assisted in the folding of a protein folding biosensor, TagRFP675 in cells⁸⁰. TagRFP675 doesn’t fold well in E. coli without co-overexpressing a chaperone (such as GroEL or DnaK). G-quadruplex forming sequences had similar protein folding activity compared to chaperone controls, demonstrating that G-quadruplexes enhanced protein folding in cells.⁷⁹ In vitro assays with chemically denatured TagRFP675 and G-quadruplexes revealed that G-quadruplexes rescued kinetically trapped intermediates to restore TagRFP675 fluorescence.⁸⁰

To further explore the mechanism behind G-quadruplex chaperone activity, two chaperone G-quadruplex sequences with solved NMR structures were systematically mutated. Single point mutations were able to dramatically affect chaperone activity in either a positive or negative sense. We found that G-quadruplex formation, stability, and topology were among the most crucial factors. We also found the ability of G-quadruplex mutants to form larger, oligomeric species correlated with chaperone activity.⁸¹

This link to oligomerization was intriguing, as our prior work had established that nucleic acids often prevent protein aggregation by facilitating protein oligomerization⁸², and that G-quadruplexes were an order of magnitude more powerful at this than bulk DNA.⁷⁹ G-quadruplexes exhibited similar oligomerization kinetics as bulk DNA, suggesting a shared mechanism of chaperone activity. As will be touched on below, this observation helps explain the frequency of observations of G-quadruplexes interacting with proteins involved in protein aggregation diseases.

Liquid-liquid phase separation (LLPS) is a phenomenon where RNA and RNA-binding proteins spontaneously form distinct liquid droplets separate from the bulk solution^83–87. Of particular interest are droplets that can form in the presence of proteins or repeat expanded RNA implicated in neurodegenerative diseases^88–101. It appears the dynamics of these droplets are highly relevant to protein homeostasis^{102, 103} and disease pathology. Although beyond the scope of this review, the roles of RNA in LLPS have been reviewed by us¹⁰⁴ and others elsewhere^{86, 92, 105}, and we would recommend these reviews to become acquainted with the current knowledge on this topic.

Nucleic Acid Roles in Protein Aggregation Diseases

Prion Protein (PrP)

One of the oldest documented dementias is induced by the misfolding of the highly conserved neuronal prion protein (PrP)^{106, 107}. PrP is a membrane-bound protein expressed across various tissues, but most highly expressed in neurons where it’s thought to play a role in myelin maintenance^{108, 109}. Prion disorders (or chronic spongiform encephalopathies) are incurable protein misfolding disorders characterized by the recruitment of healthy PrP into self-propagating, misfolded and disease-causing PrP^108–111 (Figure 4A). PrP contains two domains: a disordered, hydrophilic N-terminal domain and an ordered, hydrophobic C-terminus domain. Aggregation-prone and disease-causing mutations are concentrated in the ordered C-terminus. Converting the C-terminal domain into a β-sheet rich structure is thought to be the main driver of pathology¹¹², but aggregation of the N-terminus can also play a role^{108, 109}. In the long pursuit to determine what could cause the transition of healthy PrP to a prion form^{106, 107}, evidence began to emerge nucleic acids could modulate PrP aggregation.

Figure 4.

A general prion mechanism of PrP conversion to a prion form.

This mechanism is similar to prion-like activity seen with other proteins implicated in neurodegenerative disease. It consists of an initial event to trigger the conversion of healthy protein to a prion form yielding prion monomer and toxic oligomers. These prion forms recruit healthy protein to misfold resulting in a growing fibril. Incorporation of protein results in a plateauing of fibril growth eventually yielding protease-resistant fibrils/aggregates. Eventually, these aggregates can fragment to induce secondary nucleation events if additional healthy protein is available or added. B. A modified PMCA was used to induce prion aggregation of PrP isolated from healthy brain tissue. If the samples were treated with RNase, protease-resistant aggregates could not form suggesting the role of RNA is explicit in the prion-propagation mechanism.

Nucleic acids were often co-purified with prion disease aggregates and PrP has been shown to bind nucleic acids with nanomolar affinities¹¹³. Early work on PrP and its interaction with DNA showed that the presence of DNA was sufficient in inducing PrP fibrils^114–116. It was shown that DNA could also prevent the aggregation of PrP via the formation of β-sheet rich soluble PrP-DNA complexes¹¹³. Interestingly, these complexes were not able to seed aggregates even with excess PrP. However, in high excess of PrP aggregates, the PrP in the PrP-DNA complexes could be sequestered into growing fibrils/aggregates.

The complexities of PrP aggregation mechanisms were being uncovered, but isolating the component that templated prions remained elusive because brain extracts were likely contaminated with the templating cofactor. In an attempt to isolate the cofactor that was templating fibrils in vitro, PrP from healthy and diseased brain homogenates was treated with various nucleases¹¹⁷. They then subjected the samples to a previously used “in vitro amplification¹¹⁸” a modified protein misfolding cyclic amplification method (PMCA^{119, 120}) to yield aggregates. In PMCA, dilute samples are introduced to PrP (or other amyloid-forming protein) and allowed to aggregate. After a cycle of aggregation, the samples are sonicated, and this process is repeated to stoichiometrically recruit monomers into fibrils/aggregates. For in vitro amplification, they used dilute seeds to induce prion aggregation of healthy brain homogenates but did not periodically sonicate the samples. Instead, they just performed one extended round of seeded aggregation.

To detect if the aggregates formed were protease-resistant, they treated the post-PMCA samples with protease and used an antibody to detect PrP. They found that samples treated with ssRNA-specific nucleases were able to inhibit PMCA (Figure 4B). They also found that nuclease/heparinase-depleted samples could only yield protease-resistant aggregates if total RNA was re-introduced back into the sample. Heparin and DNA could not “rescue” the prion activity of the samples. Although compelling, it was still inconclusive what the minimum components to yield the prion conformation were¹¹⁷.

To show that RNA was an essential component of templating PrP to a prion form, soluble PrP was isolated via immunoaffinity purification¹²¹. Subsequent analysis showed the protein to be isolated in equimolar amounts with fatty acids. This was encouraging because PrP is normally membrane-bound, while aggregated PrP is cleaved from membranes. When subjecting what they assumed to be a negative control to PMCA: RNA, soluble PrP, and bound fatty acids without seeds; they yielded de novo protease-resistant aggregates of PrP. This was quite shocking, so they assumed they had used a contaminated sample. To prove they had created de novo protease-resistant PrP aggregates: they purchased separate lab equipment, bought brain samples from a different lab, and moved to a laboratory where prion proteins had never been used. The “negative” control was repeatable, and for the first time, it was shown that full-length soluble PrP could be converted to a prion form with just RNA and no existing seed¹²¹.

The broader mechanisms of soluble PrP conversion are beginning to be understood in vitro. PolyA has been shown under multiple circumstances to facilitate the conversion of PrP to a prion form^{122, 123}. It has been shown that the N-terminus is the site of nucleic acid binding and that the RNA binding of the N-terminal octarepeat can yield prion infectious aggregates¹²⁴. These nucleic acid interactions are also governed by mutations seen in disease that can enhance or ablate nucleic acid binding¹²⁵. In a computational study of mutations that cause fatal familial insomnia (FFI), it was found that mutations with enhanced nucleic acid binding could be driving the PrP to prion conformation¹²⁵. Mutations that had reduced nucleic acid binding were correlated with lower incidences of CJKD and FFI¹²⁵.

Although much work has been done to explore the prion mechanisms of PrP¹¹¹, there are still no cures or means to diagnose living patients. Aptamers have shown some promise in this regard as they can be targeted to stabilize healthy PrP or detect prion conformations^{93, 126–128}. The progression and initiation of prion disorders is still up for debate as a number of cofactors could play a role in disease pathology.

Tau

Tau is a highly expressed neuronal protein that assists microtubule assembly and flexibility that is implicated in a number of neurodegenerative disorders¹²⁹. Tau primarily exists as six isomers that can vary in both the number of N-terminal inserts and microtubule binding repeat domains (three 3R and 4R isoforms respectively)¹³⁰. These repeat domains are aggregation prone and disease-causing mutations are highly enriched in these domains¹³¹. As the association between tau aggregation and disease became clearer, researchers began to explore what influenced tau aggregation¹³². Early work by the Ginsberg lab showed that patients with Alzheimer’s Disease and other tauopathies accumulated considerable amounts of RNA in their brain aggregates^{133, 134}.

Given both DNA and RNA were known to influence microtubule assembly and bind tau, research began to explore how nucleic acids affected tau aggregation^{69, 131}. Originally, it was found that nucleic acids could modulate tau aggregation using total RNA extracts¹³⁵. In the study, both total RNA and tRNAs were able to reduce the time it took for detectable fibrils to form from 3R and 4R tau constructs from seven weeks to 14 hours. They were also able to successfully induce full-length tau (ht40, 4R, and 2N-terminal inserts) fibrils in the presence of tRNA.

The role of RNA (and other polyanionic cofactors) appears to play multiple roles in the fibrillation and aggregation of tau^136–141. The earliest studies on tau suggested RNA was seeding tau fibrillation, as the presence of RNA drastically reduced the time it took to form fibrils. Studies later showed that RNA can template tau fibrillation and act as a stabilizing scaffold for mature fibrils¹⁴². If fibrils formed in the presence of RNA are used to seed tau monomers, tau fibrillation will continue to occur so long as there is sufficient RNA. However, if RNA is consumed or not added the fibrillization will not occur. This demonstrates RNA is templating the growth of tau fibrils, as seeds from RNA templated fibrils are not sufficient on their own to induce fibrillation¹⁴².

This templating appears to occur in two different manners, RNA can be embedded at the core of tau fibrils or bind the surface. RNA-formed fibrils were incubated with fluorescently labeled heparin sulfate to determine if heparin could displace RNA. They determined RNA could be replaced by heparin sulfate in 4R tau truncations, but 3R constructs were able to retain some RNA. It has also been shown the introduction of RNase to RNA-templated fibrils causes the fibrils to dissociate into smaller fragments and soluble tau (similar effects are seen for heparin/heparinase). This indicates the structure of tau fibrils is highly variable; dependent on the cofactors present, their stability, and the tau constructs used¹⁴².

Tau fibril formation depends on the relatively disordered protein converting to a β-sheet rich structure¹⁴³. Whatever cofactor is present must also help to negate the inherent charge state of tau and orient monomers so they can be recruited into amyloid states^{136, 140, 144}. But not all RNA can do this equally. It appears tRNAs can generally induce on-pathway aggregation to yield fibrils. However, in a study exploring the effects of various cofactors, tau was found to bind polyA, U, and C with similar affinities (K_d < 10⁻⁸ M)¹⁴⁵. Only polyA was able to produce on-pathway fibrils, polyU yielded amorphous aggregates, and polyC didn’t appear to have any effect on aggregation. In a separate study, PolyG is also able to yield fibrils like polyA, but rather than the fibril core being centered around the repeat domain, it induced fibril formation on a stretch of the C-terminus¹⁴⁶.

The effects of DNA on tau fibrillation have been explored as well but the results are more conflicting. Depending on the DNA strandedness and source, DNA has been shown to have no effect or potentially induce fibrils¹⁴⁷. In one study ssDNAs were able to yield some fibril formation, but to a much lower degree than comparable RNA¹⁴⁵. The same study showed small dsDNAs were not able to yield aggregates at all. However, another study using dsDNAs showed that prokaryotic DNA was able to induce fibril formation of tau¹⁴⁸. DNA is likely also a relevant binding partner but seems to have slower binding kinetics and lower overall efficiency.

The study of tauopathies is an evolving area of research due to the number of variables involved in its aggregation. To name a few, tau interactions with RNA can yield both on and off-pathway aggregates¹⁴⁵. Tau with varying microtubule-binding repeats are not necessarily able to cross-seed another isoform¹⁴² (and one isoform is often over-represented in disease tissue¹²⁹). Solved fibril structures via cryo-EM also reveal tau fibrils formed in vitro often differ from solved structures isolated from diseased brain tissue^{149, 150}. Also, many of the interactions of tau with RNA are recapitulated with other polyanions and are dependent on the varying phosphorylation states of tau^{137, 151}. This is also ignoring the growing area of research indicating phase separation of tau with RNA and other RNA binding proteins play a role in disease⁹⁶ (see Discussion). While much has been uncovered about tau’s mechanisms of aggregation, there’s still a confluence of factors that are yet to be understood about tau pathology in vivo.

p53

Tumor suppressor protein 53 (p53) is a transcription factor that regulates a variety of genes related to DNA repair, senescence, and apoptosis. Its dysfunction can be linked to 50% of all cancers¹⁵² and it’s hypothesized the inactivation of p53 may (partly) be the result of a prion-like mechanism^{111, 153, 154}. Aggregates of p53 in cells were discovered in the 1990s and the mutations most commonly associated with disease also yield unstable, aggregation-prone species of p53^{153, 155, 156}. These destabilizing mutations also commonly occur in the core DNA-binding domain of p53, suggesting nucleic acids may play a role in its stability^{111, 157}.

In one of the earliest efforts to modulate p53 aggregation, Silva and colleagues took advantage of its nucleic binding capacity¹⁵⁸. It was widely known that p53 not only contained specific DNA binding capacity, but p53 also had an inherent affinity to non-specifically bind RNA too. To study the effects of nucleic acids on the aggregation of p53, they used the truncated core DNA binding domain of p53, p53C.

The light scattering signal of thermally denatured p53C was monitored in the presence or absence of a variety of RNAs: several regions of s23 rRNA (including domain V), tRNA, and long coding mRNAs. Generally, higher RNA concentrations were associated with lower light scattering and a delayed onset of aggregation. Of special interest, the light scattering at low RNA to protein ratios (≤ 0.3 μM RNA to 4.5 μM p53C) yielded a much higher light scattering signal than protein alone and had enhanced aggregation kinetics (i.e. a reduced lag phase)¹⁵⁸.

From here they confirmed the presence of aggregates via EM. The morphology of p53C alone and low RNA concentrations were similar but did not yield fibrils, while higher concentrations of RNA yielded no detectable aggregates. Despite not forming fibrils, they wanted to determine if the aggregates present were forming amyloid-like structures. To do so, they employed an A11 anti-body dot blot assay. A11 is an antibody specific to soluble amyloid oligomers and pre-fibrillar aggregates. They found that aggregated p53C in the absence of RNA (which yielded THT-positive aggregates) had very little A11 bound, while both low and high RNA cases yielded A11 signals. However, the higher RNA concentration had the higher dot blot signal, suggesting more soluble amyloid oligomers were formed at higher RNA concentrations¹⁵⁸.

To isolate the effects of the aggregates detected by A11, they isolated seeds from the no RNA, low RNA, and high RNA conditions. They found that seeds isolated from p53C alone or the low RNA condition had no effect on the light scattering signal compared to an unseeded reaction. Conversely, the oligomers formed at the higher RNA concentration proved to be seeding competent, enhancing aggregation kinetics and yielding a higher overall light scattering signal compared to protein alone¹⁵⁸.

They also explored the role DNA could play in the aggregation of p53C¹⁵⁹. In their earliest study, they used cognate dsDNA and a non-cognate polyGC dsDNA sequence. In the presence or absence of dsDNA, they denatured p53C via pressure and monitored its fluorescence and light scattering to determine its aggregation kinetics. They found that cognate DNA was able to enhance the stability and prevent aggregation of p53C by stabilizing the native state. Cognate DNA was also able to rescue p53C that was previously denatured in the absence of DNA. They found that non-cognate DNA, polyGC, was able to prevent overall aggregation of p53C but it did not bind or stabilize a native state. Therefore, unlike RNA that initially appeared to largely aggravate p53 aggregation, that specific binding to its cognate DNA prevented aggregation and stabilized the native state¹⁵⁹. In a similar observation, an RNA aptamer was designed against a p53 aggregation-prone mutant, p53 R175H¹⁶⁰. Shan and colleagues showed that the peptide was able to specifically target the disease-relevant mutant and that the aptamer was able to rescue p53 function in cells¹⁶⁰. The aptamer also proved effective at shrinking tumors caused by mutated p53 in a mouse model. Therefore, specific binding of p53 could play an important role in stabilizing it.

However, the case of DNA vs RNA in p53 was not so simple. In follow-up work with DNA, they explored the concentration dependence and non-specific interactions of DNA on the aggregation of p53C. When they thermally denatured p53C in the presence or absence of non-cognate dsDNA, they found that the light scattering signal of p53C increased with increasing dsDNA concentrations. In contrast, the ssDNA yielded a similar pattern to RNA where light scattering decreased with ssDNA concentration (albeit with much less potency)¹⁵⁸. From these studies, it is clear that at least for p53, the difference in the effects on aggregation and folding between specific and non-specific nucleic acid interactions with proteins can be large.

α-synuclein

A highly expressed protein in neurons, ⍺-synuclein is a small, (14 kDa) intrinsically disordered protein that adopts an α-helical rich structure upon membrane binding^{161, 162}. It plays a role in neurotransmitter release as well as DNA repair mechanisms¹⁶³, but the various functions of α-synuclein are still an active area of research^{161, 163}. The aggregation of α-synuclein is associated with neurotoxicity, eventually leading to neurodegeneration, with the protein playing an outsized role in Parkinson’s disease (PD) and other dementias^{164, 165}. Early work focused on the fibrillation of α-syn to discern possible mechanisms behind PD pathology¹⁶⁴. This work led to the discovery that nucleic acids could also modulate the aggregation of α-synuclein.

Previous work had shown histones were able to template α-syn fibrillation¹⁶⁶. To determine if DNA was also playing a role in this process, ⍺-syn was incubated with dsDNA and aggregation was monitored by THT¹⁶⁷. Compared to α-syn alone, DNA greatly reduced the lag phase of fibril formation for WT-⍺-syn, cutting the t_1/2 from 150 to 50 hours. These aggregates were confirmed to be forming fibrils by EM. In this same study, dsDNA was also able to nucleate fibril formation for inherited PD disease-relevant mutants A50T and A30P.

Work expanding the mechanism behind ⍺-syn aggregation showed that stabilizing certain conformations was beneficial to delaying or inducing its fibrillation¹⁶⁸. Generally, stabilizing an α-helical or its native random coil species resulted in delayed aggregation, while β-sheet conformations enhanced aggregation. In the presence of DNA, the conformation of α-syn appears to be dependent on the structure and sequence of DNA present. Supercoiled DNA caused α-syn to adopt both α-helical and β-sheet conformations, ssDNA appeared to bind weakly or induce β-sheet conformation, and circular ssDNA induced an α-helical conformation. In an aggregation assay where fibril formation was detected by THT, these conformations appeared to be correlated to the kinetics of fibril formation. The ssDNA enhanced fibril formation, similar to previous work, yielding a t_1/2 of 5 hours compared to 15 hours for protein alone. Both supercoiled DNA and circular ssDNA delayed fibril formation with half-lives of approximately 45 and 60 hours respectively.

The above study also yielded a mild correlation between α-syn conformation and sequence. Calf thymus DNA is relatively low in GC content compared to lambda DNA (42% vs 70%), and the GC rich-lambda DNA yielded more β-sheet conformation whereas calf thymus DNA yielded little conformational change. It appears that both nucleic acid sequence and structure are important factors in α-syn’s aggregation, aptamers are uniquely designed to take advantage of this.

Aptamers by their nature come from a large pool (>10¹⁰) that can sample both conformational and sequential spaces¹⁶⁹. Due to its high degree of flexibility, α-syn is a difficult protein target for drug design. An aptamer that binds may or may not have the intended effect, or the aptamer may specifically bind a lowly populated state of the protein in vivo. The first aptamers for α-syn were designed in the early 2010s but appeared to have little effect on the aggregation pathway of α-syn¹⁷⁰. Work began to emerge that the aggregation of α-syn could be modulated by aptamers from several labs in the following years^171–174. Chen and colleagues discovered aptamers that were able to bind oligomerized α-syn and enhance off-pathway amorphous aggregation¹⁷². In some of the more promising PD work to date, Zhang and colleagues found two aptamers that were able to prevent α-syn fibrillation via Thioflavin T (THT) binding assays and EM¹⁷⁴. The aptamers were effective in preventing α-syn aggregation in neuroblastoma cells, rescuing mitochondrial dysfunction, and enhancing neuronal morphology compared to random DNA. They were also effective in rescuing PD-model mice at both the cellular and behavioral level¹⁷¹.

Several of these aptamers for α-synuclein appear to contain G-quadruplex forming sequences. It was recently shown that G-quadruplexes can tightly bind to α-syn fibrils and that the α-syn fibers can change the conformation of the G-quadruplexes¹⁷⁵ In a recently published pre-print, it was shown α-syn can bind G-quadruplexes via it’s N-terminus. They also treated mice PD models by destabilizing RNA G-quadruplexes shown to enhance α-syn aggregation⁸⁸. This remains an important area for future study.

Huntingtin protein (Htt)

Huntington’s (HD) disease is a genetic neurodegenerative disorder caused by the Huntingtin protein. Huntingtin protein (Htt) is a relatively large (350 kDa) ubiquitously expressed protein that is most concentrated in the brain^176–178. Htt contains an aggregation-prone polyQ region where the number of repeats is highly associated with Huntington’s disease and increased aggregation propensity^176–178. Early efforts to treat the disease focused on targeting this aggregation-prone region^{176, 179}. This led to the discovery that oligonucleotides and chemically modified oligonucleotides could prevent Htt aggregation and cell death.

Htt is known to bind its first exon, so the Kmeic group screened sequences end capped by phosphorothioate linkages in the sense and anti-sense direction of the gene. Using sequences that ranged from 53 to 6 oligonucleotides in length, they found that regardless of the oligo length, modified ssDNA was able to prevent mut-Htt aggregation and enhance cell viability¹⁸⁰.

In follow-up work from the Kmiec lab, they found G-quadruplex forming sequences were as effective as the original modified nucleic acids at preventing mut-Htt aggregation. An all-guanine 20mer yielded similar aggregation levels to the previous best candidate prevention at a 40x lower concentration. In contrast, poly A, C, and T 20mers made aggregation worse, yielding 1.5 to 2.5x more aggregates than the mut-Htt alone control. The all G-20mer was also able to enhance cell viability in cells expressing mut-Htt compared to untreated and mock treatment controls¹⁷⁹.

The Kmiec group and others found that these oligonucleotide sequences were binding specifically to the polyQ regions^{101, 181}, leading to the development of aptamers for mut-Htt. These aptamers have had success rescuing mitochondrial dysfunction, cell viability, and improving neuron function in a D. melanogaster HD model¹⁸². It is an active area of research and an optimistic route for treating HD^183–188, but it largely remains to be explored whether G-quadruplexes could also have a pathological role in HD.

Ataxin-1

Spinocerebellar ataxias (SCAs) are a group of diseases characterized by repeat expansions, similar to conditions such as HD and Frontotemporal Dementia/Amyotrophic Lateral Sclerosis (FTD/ALS) associated with c9orf72 expansions^189–191. The number of nucleotides involved in a specific SCA type varies, with tri-, penta-, and hexanucleotide repeats observed in clinical cases. These repeats can occur in coding or non-coding regions of a gene. It’s worth noting that the severity, onset, and progression of SCAs are linked to the number of repeats present^189–191.

SCA type 1 (SCA1), the first well-characterized SCA, results from CAG repeats in the ATXN1 gene, leading to a polyglutamine expansion in the ataxin-1 protein^192–194. Ataxin-1 is primarily found in the nucleus of nervous tissue but is also expressed throughout the body. Like other polyglutamine expansions, ataxin-1 with a polyglutamine expansion (ataxin-1[polyQ]) tends to aggregate, with 39 polyQ repeats considered a diseased state^{193, 195}. Ataxin-1 is a gene repressor and functions primarily as a DNA-binding protein, although it can also interact with RNA^{196, 197}.

Studies using ataxin-1 with varying polyQ repeat lengths have demonstrated its ability to bind both polyU and polyG RNA, with a stronger affinity for polyG¹⁹⁶. Notably, the affinity of ataxin-1 for RNA seems to depend on the size of the polyQ expansion. Ataxin-1 lacking glutamine repeats showed the strongest RNA affinity, followed by ataxin-1[30Q] and ataxin-1[82Q]¹⁹⁶. This is intriguing, considering RNA’s role in modulating the aggregation of other polyglutamine expansion disorders.

Early studies in mice suggested that the nuclear localization of ataxin-1, rather than nuclear aggregation, was crucial to disease pathology in an SCA-1 mouse model¹⁹³. Subsequent research in cells demonstrated that ataxin-1 with 2, 26, and 80 glutamine repeats localized to nuclear punctate, a process dependent on the presence of RNA¹⁹⁷. Inhibition of transcription and RNase treatment reduced nuclear punctate for ataxin-1, which was also found to co-localize with RNA in the nucleus. Notably, normal ataxin-1 could efficiently shuttle in and out of the nucleus, while ataxin-1 with an 80Q expansion could not be exported to the nucleus, echoing the findings of earlier mouse studies¹⁹⁷.

Despite tempting links to other repeat expansion disorders like HD, there remains controversy surrounding the mechanisms of many SCA disorders^{189, 190}. For SCA-1, questions persist regarding whether the polyQ expansion can exit the nucleus and what drives disease pathology^{193, 195–197}. Other SCAs associated with ataxin polyQ expansion disorders also may not necessarily share the same disease mechanisms^{198, 199}. However, despite the complexity, investigating the role of RNA in the aggregation of SCA-related proteins offers a promising avenue for further research and potential treatments.

Conclusions

The involvement of nucleic acids in protein aggregation is an emerging area of research. Nucleic acids can aid in both protein folding and preventing aggregation. Conversely, nucleic acids can also play an antagonistic role in promoting protein aggregation. This activity appears to be inherent to polyanions, however, nucleic acids are much more potent in exerting this activity compared to its backbone alone (polyphosphate). This indicates there are unique structural and sequential features of nucleic acids that drive these behaviors.

Much focus has been given to the potential roles nucleic acids play in disease pathology (see Table 1), however, there is an exciting potential their activity can be exploited in biotechnology. It has already been shown that tRNA can provide chaperone function for the folding and assembly of monomers into virus-like particles in vaccine technology²⁰⁰.

Table 1:

Disease-related proteins and their interactions with nucleic acids in aggregation discussed in this article.

Protein	Nucleic Acid Involvement	References
prion protein (PrP)	RNA can induce de novo formation of the prion form of PrP, resulting in on-pathway aggregation.	93, 106–111, 113–117, 121–128
Tau	RNA is able to convert Tau into on-pathway conformations, oligomers, and stabilize amyloid structures. Other polyanions such as heparin sulfate can also yield amyloid and on-pathway oligomerization.	69, 96, 129–151
Tumor suppressor protein 53 (p53)	RNA can induce on pathway or amyloid containing/pre-fibrillar aggregates of p53 truncations. Cognate DNA can stabilize the native state, but non-cognate DNA has less consensus. Aptamers have shown success in rescuing p53 activity.	111, 153–160
⍺-synuclein	The aggregation of ⍺-synuclein is sensitive to the structure and sequence of nucleic present. Aptamers can rescue ⍺-synuclein, G-quadruplexes can induce aggregation, and various DNA can reduce the lag phase of aggregation.	88, 161–175
Huntingtin protein (Htt)	Phosphorothioate-modified, G-quadruplex, and polyG DNA can prevent protein Htt aggregation. Other polynucleotide sequences appear to exacerbate Htt aggregation. Aptamers have shown promise in treating disease models.	101, 176–188
Ataxin-1	Ataxin-1 nuclear localization and puncta formation depend on the presence of RNA, which may be relevant in the disease pathology.	189–199

Another intriguing area of research is the use of aptamers to target disease-relevant proteins to potentially prevent their aggregation and oligomerization. Further research in this area has the potential to provide insights into the mechanisms underlying these devastating conditions and potential avenues to treating a variety of diseases.

Our group and others have found that G-quadruplexes are particularly adept at chaperoning protein clients and interacting with many disease-relevant proteins. Furthermore, there are client-specific preferences for nucleic acid interactions, as observed with p53, ⍺-synuclein, Htt, and others. These interactions appear to be highly nuanced as it can be difficult to predict the behavior of the client protein (i.e. whether it will aggregate, refold, etc.) with a given nucleic acid.

While some sequence and structure-specific information on RNA in protein folding and aggregation has been discovered, we speculate that many more RNAs in particular will be involved in important ways. As an analogy to studies of proteins, each protein is assumed to have a unique function based on its structure. Many who study proteins do not assume this level of structure-function relationship in RNA, but we believe this is erroneous. Ultimately, there may be various RNA structures with differing effects on protein folding and aggregation. The long-term future of this research field depends on the ongoing discovery of diverse structures and the determination of biophysical rules governing these structure-function relationships.