Emerging Roles for G-Quadruplexes in Proteostasis

Abstract

How nucleic acids interact with proteins, and how they affect protein folding, aggregation, and misfolding is a still-evolving area of research. Considerable effort is now focusing on a particular structure of RNA and DNA, G-quadruplexes, and their role in protein homeostasis and disease. In this State-of-the-Art Review, we track recent reports on how G-quadruplexes influence protein aggregation, proteolysis, phase separation, and protein misfolding diseases, and pose currently unanswered questions in the advance of this scientific field.

Introduction

Protein homeostasis, also called proteostasis, is vital for cellular health. The proteostasis network consists of chaperones that aid in protein folding or prevent aggregation, as well as the protein degradation machinery [1]. Together, the proteostasis network sustains a favorable protein environment preventing the accumulation of potentially toxic protein aggregates that are often associated with protein misfolding diseases [2]. This view of protein folding and aggregation is, unsurprisingly, protein-centric. However, considerable data now suggest that nucleic acids also have an important role in protein folding and aggregation.

This review is not intended to be a comprehensive review of the literature on how nucleic acids affect protein folding. Instead, as a State-of-the-Art review, we focus on recent literature and project into the near future as to possible avenues that the study of nucleic acids in protein folding could take. This review will primarily focus on the potential future of G-quadruplex nucleic acids and their role in proteostasis, and discuss how RNA is an important player in protein solubility in the cell.

Nucleic acids as protein aggregation and folding modulators

Experiments examining how nucleic acids can affect protein folding were originally performed by Das Gupta et al. and expanded by Sanyal et al. in a series of papers examining the ribosome as a molecular chaperone, especially Domain V of the 23S rRNA [3–6]. This work has been reviewed elsewhere [7, 8], and will not be repeated here. Similarly, other groups have continued to probe the chaperone effects of various RNAs, especially in vitro [9, 10]. More recently, we found that a large swath of nucleic acids have powerful chaperone activity, suggesting that they could, in bulk, play a significant role in protein solubility in the cell, and that even simple poly-U RNA could potentially aid protein folding indirectly [11, 12]. This concept garnered further support in a recent study in which Aarum et al. carefully tracked the effects of adding RNase to cellular lysates. RNase treatment resulted in massive protein aggregation, and not just of known RNA binding proteins [13]. These findings highlight the importance of RNA in preserving a favorable protein folding environment, but which DNA or RNA sequences and/or structures are most important in modulating protein folding and aggregation remains an open question.

G-quadruplex structure

Quadruplexes are a diverse set of nucleic acid structures. Although other forms of quadruplexes can be formed, guanine-quadruplexes are the best understood. The structural basis of the G-quadruplex is a base tetrad, in which four guanine bases form Hoogsteen basepairing in a planar formation, with a monovalent metal ion running down the center of the tetrad. The stacking of multiple tetrads provides additional structural stability. From this starting point, G-quadruplexes can take on a remarkable number of structural topologies and can be mixed with other nucleic acid structural elements, such as helices, loops, and bulges [14–16]. G-quadruplexes are typically classified by the topology of their backbone directionality. In parallel G-quadruplexes, all four backbones run in the same direction with respect to each other in the structure. In anti-parallel G-quadruplexes, the direction of two of the backbones runs in the opposite direction. Finally, in 3+1 mixed structures, three of the backbones run parallel, with a single strand running anti-parallel [15]. DNA G-quadruplexes are now well established to exist in cells, and RNA G-quadruplexes have also been detected in a variety of different contexts, although given the dynamic nature of RNA prevalence in the cell is less clear [16–18]. Previously, G-quadruplexes have been shown to play roles in regulating transcription, translation, splicing, and localization [19].

G-quadruplex-Protein Interactions

In a simple model of chaperone activity, RNA G-quadruplexes would directly, or perhaps indirectly, complex with target factors to promote protein folding or prevent aggregation. Proteins known to bind RNA G-quadruplexes are largely classified as nucleic acid binding proteins harboring large stretches of protein disorder. Early examples of G-quadruplex-protein interactions were described at telomeric regions in ciliated protozoans, where telomere end-binding proteins were found to interact with G-rich DNA sequences presumed and later demonstrated to fold into G-quadruplexes [20, 21]. Since then, many additional candidate G-quadruplex binders have been identified and characterized in vitro and in vivo.

A mounting body of literature indicates that G-quadruplexes bind disordered protein domains with numerous in vitro studies demonstrating direct high-affinity binding, previously reviewed in [22] and [23]. In at least two cases, the structure of the G-quadruplex-protein complex has been solved by X-ray crystallography [24, 25]; however, a general mechanism underlying these interactions remain largely unclear. Techniques such as NMR spectroscopy, surface plasmon resonance, and electrophoretic mobility shift assays, coupled with mutagenesis in either the RNA or protein, have proven useful for understanding protein-G-quadruplex interactions. Huang et al., for example, describe the characterization of the G-quadruplex binding protein, cold-inducible RNA-binding protein (CIRBP), utilizing the aforementioned techniques [26]. In addition to targeted assays, several studies have employed unbiased approaches to explore and validate G-quadruplex binders. Goering et al. utilized a large-scale in vitro assay, RNA bind-n-seq, to validate FMRP-RNA-G-quadruplex interactions with similar success to SELEX-based approaches [27, 28]. Herdy and collaborators used biotinylated RNA G-quadruplex to affinity capture interacting proteins from cellular lysates [29]. This study and similar proteomic approaches have primarily identified proteins with RNA processing related activities and regions of disorder, including DDX3X, FMR1, NCL and many others [30–32]. However, at least one of these large-scale studies has also captured heat shock family proteins (e.g. DNAJC9 and HSPA5) [32]; whether these interactions are direct, occur in intact cells, and have biological significance remains to be determined.

G-Quadruplexes in Modulating Aggregation

Although previous work had tested various nucleic acids for chaperone activity in vitro [33], many of these effects were difficult to test for biological relevance. This was especially true of the ribosomal RNAs, which are difficult to test for in-cell activity due to the challenges of mutagenizing to disrupt chaperone activity without impacting protein synthesis. Therefore, in previous work, we set out to determine in a more unbiased fashion what sequences produced optimal chaperone activity via an in vitro screen.

We devised a screen to test for aggregation prevention activity, as this was the activity we had previously observed for nucleic acids. The results of the screen were quite clear: among short nucleic acids, by far the best at preventing protein aggregation were G-quadruplex-containing oligonucleotides. The best G-quadruplexes were over an order of magnitude more effective at preventing aggregation than bulk DNA. Subsequent testing showed that there is a topological dependence to the activity, with anti-parallel G-quadruplexes performing the worst at chaperone activity, 3+1 mixed G-quadruplexes performing the best, and parallel G-quadruplexes in the middle. We also found that at least part of the chaperone activity stemmed from promoting kinetically stable oligomers containing protein and nucleic acid components, and that this oligomerization activity was far more potent than that of bulk DNA or ssDNA. Summarizing the in vitro data, we found that G-quadruplexes were powerful modulators of protein aggregation and oligomerization, with the potential to govern both. We additionally tested several of the G-quadruplexes for their ability to improve the protein folding environment of E. coli, and found that they improve the folding of an unstable fluorescent protein in E. coli [34].

Combined with the predilection for G-quadruplexes to bind disordered proteins, the strength of G-quadruplex activity at preventing protein aggregation and promoting oligomerization lays the foundation of a hypothesis that G-quadruplexes are important players in proteostasis. However, there has also been considerable work in the very recent past that suggests other possible roles. These mechanisms can be broken into three categories: 1) G-quadruplexes as managers of phase separated states, 2) G-quadruplexes in protein misfolding diseases, and 3) the role of G-quadruplexes in proteolysis and their emerging presence in the ribosome. The current literature and unanswered questions will be addressed in each area.

G-quadruplexes participating in LLPS and phase separated states

LLPS has gained attention recently as an important mechanism of organization in the cell. In cells, the composition of these LLPS (also referred to as condensates) is largely a mixture of RNA and proteins and can occur in the nucleus and the cytoplasm and are often termed granules. The role of RNA in these granules is still an active area of research, but it is already clear that RNA plays an important role in promoting or inhibiting phase separated bodies and maintaining their fluidity [35]. The best studied of these granules are stress granules (SG), which form as a function of protein folding stress, and are thought to serve as a holding place for proteins, nucleic acids, and ribosomes until stress has passed [36]. In addition to the important role of proteins in forming these stress granules, mRNAs have recently emerged as important co-factors in shuttling proteins to the granules at the onset of stress [37, 38].

What are the roles of G-quadruplexes in LLPS? A set of recent studies have begun to give us clues (Fig 1). Although the vast majority of known LLPS events in cells likely consist of diverse and complex mixtures of nucleic acids and proteins, the relative importance of nucleic acids in this process can be assisted by analyzing simple systems with individual nucleic acids. For example, Zhang and colleagues performed a study of the SHORT ROOT (SHR) mRNA [39], which is important for Arabidopsis root development [40, 41]. After showing that SHR phase separates in plant cells, the authors performed in vitro experiments to further decipher mechanism. Their findings indicated that the primary trigger for phase separation of SHR is a G-quadruplex-forming segment, and that the properties of this G-quadruplex govern the properties of phase separation of the mRNA on its own [39].

Figure 1.

Source of Quadruplexes in Liquid-Liquid Phase Separation. Both A. mtDNA cleaved under oxidative stress in G-rich stretches and exported from the organelle and B. SHORT ROOT (SHR) mRNA or quadruplex rich segments of mRNA, often found in 5’-UTR’s, can phase separate solely in the presence of nucleic acids and/or in the presence of proteins. C. tRNA under oxidative stress conditions is cleaved to form tiRNA. Quadruplex forming tiRNA is bound by YB1 and the complex can go on to induce phase separation with nucleic acids and other proteins.

In simplified protein:RNA mixtures, the presence of a G-quadruplex can initiate phase separation. Polynucleotide oligos (poly-rU, poly-rA, and poly-rC) induce spherical LLPS of poly-arginine peptides. Poly-rG, however, induces the formation of fractal-like states that are still dynamic like an LLPS state, suggesting that poly-rG can induce higher levels of order while still retaining a fluid state [42, 43]. Even RNA alone can undergo phase separation, of note, G-quadruplex and disease-associated repeat RNAs formed large assemblies in cells and in vitro [44]. Moving to full proteins, there are also now multiple reports in which G-quadruplexes trigger the phase separation of single protein:G-quadruplex mixtures, such as the linker histone H1 [45]. In the case of H1, the work was in vitro, but the authors were able to correlate the level of G-quadruplex formation with the amount of LLPS formed.

How might these effects play out in biological observations? As a steppingstone, Liu et al. created a simplified cellular environment using protocells from giant membrane vesicles to test the LLPS possibilities of G-quadruplexes. In these protocells, G-quadruplexes were found to phase separate with a largely disordered protein SERBP1 [46]. In another study, Gao et al. found that the helicase Ded1 bound to but did not unwind G-quadruplexes. Instead, the authors observed that G-quadruplexes-Ded1 interactions led to LLPS. In this case, they were able to extend these results by introducing the G-quadruplex stabilizing small molecule Phen-DC3 into cells with fluorescently labeled Ded1 [47]. The addition of Phen-DC3 caused formation of a spherical membraneless organelle containing Ded1, strongly suggesting that the Ded1:G-quadruplex interaction and the RNA structure were responsible for its formation.

Finally, G-quadruplexes participate in the creation of stress granules. During oxidative stress, Byrd and colleagues found that G-quadruplexes released from mitochondrial DNA, migrate to the cytoplasm and organize into stress granules [48]. In a more indirect mechanism, tRNAs are often cleaved to create tiRNAs, which can repress translation at the onset of stress [49]. These tiRNAs then form G-quadruplex structures, and trigger stress granule formation via binding cold-shock protein YB-1 [50]. It should be noted that this case is different from the previous cases in that it currently is not hypothesized that tiRNA G-quadruplexes affect stress granule formation directly [50].

In a related case, He et al. analyzed and performed crosslinking and immunoprecipitation (CLIP) followed by RNA sequencing to demonstrate RNA G-quadruplex-binding by the stress granule associated protein, G3BP1 [51]. Consistent with previous work, a disordered region of G3BP1 was critical for the interaction and normal protein function [51]. It should be noted that G3BP1 is a major stress-granule component that has been shown to modulate their assembly in a manner that depends on its disordered domains and RNA [52, 53].

Within this realm of study, there are still many open questions regarding how G-quadruplexes influence LLPS and granule formation, including mechanistic and biological aspects. What appears to be most likely based on the trajectory of recent studies is that G-quadruplexes are potent at inducing phase separated states, which would appear to be a natural consequence of their ability to promote protein oligomerization. Finally, while much of the work on phase separation to date has relied on assessing simple in vitro systems often involving a single protein, the cellular environment is a complex mixture of macromolecules and metabolites. Thus, high affinity interactions between proteins and G-quadruplexes may provide a path for seeding phase separation and organization in this complex environment

The Role of G-quadruplexes in Proteolysis and Protein Folding

In addition to aggregation prevention, other standard functions of the proteostasis network are to regulate protein degradation and folding. Although the work on G-quadruplexes in this area is still highly speculative, there are tantalizing clues leading to unanswered questions.

On the side of proteolysis, nucleic acids have been known for years to interact with the protease Lon from both E. coli and mitochondria [54, 55]. The effects of nucleic acid binding have been primarily observed to be stimulatory in the case of E. coli Lon but having mixed effects for mitochondrial Lon [55–58]. It was recently found that in addition to other nucleic acids, Lon binds G-quadruplexes [55] (Fig 2), and very recent data point to this being a preferential binding partner [59]. While studies of the effects of Lon:G-quadruplex binding are still ongoing, early evidence points to G-quadruplexes having an especially strong effect on Lon activity [59]. How this effect on Lon activity interacts with the chaperoning effects of G-quadruplexes is still an open question.

Figure 2:

Lon protease interacting with G-quadruplexes, with the G-quadruplexes affecting protein degradation.

Much less is currently known on the role of G-quadruplexes in protein folding. Previous data for other nucleic acids suggest that at least a passive mechanism for improving protein folding is possible, but no direct links have yet been established. Many protein chaperones are known to interact with nucleic acids [11, 12, 60, 61], but their interactions with G-quadruplexes have not been well-explored to date. One tantalizing possibility stems from the recent finding that rRNA extensions (or tentacles), which extend away from the ribosome, are highly enriched in G-quadruplexes [62–64]. This observation puts a potentially powerful influencer of protein folding in close proximity to newly translated proteins. Given the ability for G-quadruplexes to modulate aggregation, LLPS, and proteolysis, continued exploration of the role of G-quadruplexes in protein folding should be of high priority.

G-quadruplexes in Protein Misfolding and Aggregation Disease

Nucleic acids have long been known to affect prion and amyloid formation [65–69]. Based on the ability for G-quadruplexes to modulate protein aggregation and LLPS, it follows that there are an increasing number of reports of G-quadruplexes interacting with proteins implicated in protein misfolding diseases. Here we will focus on three disease cases: Amyotrophic lateral sclerosis (ALS), Fragile X Syndrome, and Alzheimer’s Disease.

ALS

ALS is marked by degeneration of motor neurons, which causes patient paralysis within a few years of diagnosis. Far from having effective cures and preventions, even the cause of ALS is still heavily debated [70]. Part of this debate stems from the wide range of genetic variables that can seemingly cause ALS [70]. However, there are a few hallmarks of the disease at the biochemical level. Among these are mutations thought to trigger protein aggregation, such as SOD1 and especially TDP-43 [70], and overexpression of the RNA C9ORF72, which harbors a repeat expansion consisting of GGGGCC repeats [70]. Recent theories for the disease onset point to the possibility that a driver of the disease is the formation of stress granule-like formations that persist and worsen with disease progression [71, 72]. This chronic stress granule formation could potentially have several different biochemical cues and thus could account for the sporadic and familial cases of ALS [73].

As discussed above, G-quadruplexes are powerful mediators of LLPS and aggregation, so they seem a likely candidate to be involved in ALS. And indeed, over the past several years studies have found that many of the proteins underlying ALS are also G-quadruplex binding proteins, including TDP-43, FUS, Ewing’s sarcoma protein EWSR1, hnRNP A2/B1, hnRNP A3, and TIA1 [74–82] (Fig 3). Furthermore, C9ORF72’s GGGGCC repeats themselves form G-quadruplexes [83, 84]. It is therefore intriguing to ponder whether it would be worthwhile to re-frame ALS as a disease of G-quadruplexes and its binding partners in the search for mechanistic causes and potential treatments. For FUS in particular, Ishiguro and colleagues showed, in vitro, that disease-linked FUS binds RNA G-quadruplexes, and this interaction leads to an increase in the propensity of FUS to form condensates. Furthermore, Ishiguro noted that the G-quadruplex-FUS condensates transition to solid-like aggregates [85]. Thus, driving forces behind ALS in large part include disordered proteins that bind G-quadruplexes or toxic RNAs that themselves form G-quadruplexes.

Figure 3:

Examples of partially disordered proteins known to interact with G-quadruplexes and aggregate in ALS. Surface representation of positive patches (in blue) and negative patches (in red) show widespread surfaces in structured regions that could electrostatically enhance nucleic acid binding. PDB codes: 6FN8 (SOD1), 6T4B (TDP-43), 2LA6 (FUS), 2CPE (EWSR1), 5WWF (hnRNP A2/B1), 5O3J (TIA1).

Fragile X Syndrome

Initially, Fragile X Syndrome was found to be caused by loss of the FMRP protein via an expanded CGG repeat that ultimately silences the FMR1 gene [86]. Later, it was found that mutations in FMRP could also cause the disease, presumably by promoting instability and protein aggregation/degradation [87]. In its natural role, FMRP is known to bind G-quadruplex RNAs and facilitates RNA transport [28]. Given the possibility of G-quadruplexes to affect protein oligomerization, studying how G-quadruplexes specifically affect FMRP aggregation and function is also an important possible avenue for disease treatment. As mentioned earlier, Darnell et al. showed RNA G-quadruplex binding by the disordered region of FMRP [27]. Since then, additional in vivo and in vitro experiments have supported this direct interaction. A crystal structure of the interaction has been reported, but it remains possible that many other binding modes between FMRP and G-quadruplexes also exist [25]. Functionally, Goering et al. demonstrated that FMRP-G-quadruplex interactions promote localization of G-quadruplex-containing mRNAs to neuronal projections. Importantly, FMRP-patient derived neurons show mis-regulation of rG4-containing FMRP targets [28].

Alzheimer’s Disease

Some of the first evidence that nucleic acids could affect protein aggregation in disease came from staining of Alzheimer’s Disease patient aggregates over two decades ago [88, 89]. These studies found high amounts of nucleic acids in both the plaques and tangles of Alzheimer’s Disease. However, it was only very recently that a study was performed to identify which nucleic acids were present in these disease aggregates. Reis and colleagues performed sequencing on aggregates from both nuclear and cytoplasmic aggregates from multiple Alzheimer’s Disease patients. The authors were able to identify many RNAs and DNAs that were enriched in the cytoplasmic and nuclear aggregates, respectively. However, a theme in both cases was that there was a high enrichment in poly-G containing sequences that were predicted to form G-quadruplexes [90]. The high enrichment of potential G-quadruplex forming sequences in Alzheimer Disease aggregates is consistent with their roles in LLPS and in modulating protein aggregation.

Cancer

In addition to the roles of G-quadruplexes in the aforementioned diseases, both DNA and RNA G-quadruplexes have emerged as structures underlying cancer-associated phenotypes. While a direct link to proteostasis in these cases is less clear, DNA and RNA G-quadruplexes underly altered transcription, splicing, and translation of oncogenes and tumor suppressors (reviewed in [91]). Indeed, some of the best biochemically and functionally characterized G-quadruplexes are found in the promoters (DNA) and untranslated regions (RNA) of cancer-associated transcripts (e.g. NRAS, MYC, and BCL2) [19]. Specific to proteostasis, the frequently mutated p53 protein is known to aggregate in cancer, and this aggregation can be dictated by nucleic acid binding or loss thereof [92, 93]. NPM1 also aggregates in cancer and directly binds G-quadruplexes, although these two aspects have not yet been tested for codependency [94, 95]. These examples suggest a potential role for G-quadruplexes in modulating protein aggregation in cancer. However, how G-quadruplexes might directly impact these proteins or cancer associated phenotypes remains to be elucidated.

Outlook and Major Outlying Questions

As discussed here, the field of G-quadruplexes in proteostasis is rapidly evolving, with many studies emerging only within the past year. As such, there are many open questions. Answering some of these questions are difficult and will require a continued push in the technology of identifying G-quadruplexes and their interacting partners, especially in cells. Being able to accurately predict and experimentally identify G-quadruplexes from sequences and sequencing of RNA in protein aggregates will be essential.

It should be noted that G-quadruplexes are not the only form of quadruplexes. For example, in their study of linker histones and G-quadruplexes touched on earlier, Mimura et al. also tested whether i-motif quadruplexes formed of cytosines could also cause LLPS, and the answer was yes [45], demonstrating that G-quadruplexes may only be the beginning. However, it is still important to note that the incredible structure variation of G-quadruplexes alone could yield a tremendous amount of variation in effects. For example, as mentioned above, we found that different G-quadruplex topologies have different levels of ability to prevent protein aggregation [34]. This was a small sampling of the possible structural space of G-quadruplexes and continuing to explore the relationship of the structure and dynamics of G-quadruplexes to their effects on protein folding and aggregation will likely continue to remain a fruitful area of research.

A pressing set of questions in this field are the extent to which RNA G-quadruplexes are folded in cells [17], the impact their structure has on protein binding, and their emerging role in proteostasis. Literature in this area is growing, with several studies using genomic approaches having identified transcriptome-wide G-quadruplex-protein interactions for numerous factors. In some cases, biochemical data indicate binding is disrupted in conditions that disfavor G-quadruplex folding [25, 28, 96]. These studies suggest that binding sites in cells exist(ed) in a folded state to mediate the interaction. Furthermore, helicases with an apparent affinity and specificity for G-quadruplex unfolding have also been characterized (e.g. DHX36), indicating that their activity relies on G-quadruplex existing in a folded state [97, 98].

Directly impacting how we interpret the role of G-quadruplexes in modulating protein stress, a recent preprint by Kharel et al. revealed RNA G-quadruplex folding was enhanced under conditions of cellular stress [99]. Furthermore, this stress-induced folding was reversible after the stress subsided. The same study also demonstrated that G-quadruplex unfolding is an ATP-dependent process, supporting that helicases have evolved to act on these structures [99]. Thus, G-quadruplexes are most likely to be formed under the conditions in which protein oligomerization and aggregation is at its highest, and the potential for modulating these processes is the greatest.

In addition, many other questions of interest remain. For example, what is the role of G-quadruplexes in autophagy? It was recently shown that G-quadruplexes can regulate autophagy [100], but do those G-quadruplexes have direct action on modulating protein fate? Other pressing questions include: What happens to the folding and stability of disorder-containing proteins that bind G-quadruplexes if the interaction is prevented? Primary direct G-quadruplex binders are often disordered; could their G-quadruplex interactions stabilize them? Which interactions between G-quadruplexes and proteins are functional in proteostasis? Do assemblies involving LLPS promoted by RNA G-quadruplexes favor folding or protein stability? These and many other questions remain open for future investigation.

Abbreviations:

CIRBP

cold-inducible RNA-binding protein

ssDNA

single-stranded DNA

LLPS

liquid-liquid phase separation