Synthetic biology tools to promote the folding and function of RNA aptamers in mammalian cells

ABSTRACT

RNA aptamers are structured RNAs that can bind to diverse ligands, including proteins, metabolites, and other small molecules. RNA aptamers are widely used as in vitro affinity reagents. However, RNA aptamers have not been highly successful as bioactive intracellular molecules that can bind target molecules and influence cellular processes. We describe how poor RNA aptamer expression and especially poor RNA aptamer folding have limited the use of RNA aptamers in RNA synthetic biology applications. We discuss innovative new approaches that promote RNA aptamer folding in living cells and how these approaches have improved the function of aptamers in mammalian cells. These new approaches are making RNA aptamer-based synthetic biology and RNA aptamer therapeutic applications much more achievable.

Keywords:

• RNA folding
• RNA aptamer
• SELEX
• misfolding
• molecular chaperone
• RNA scaffold
• RNA sensor

Introduction

Many emerging RNA synthetic biology applications involve expressing structured RNAs in mammalian cells to modulate cellular processes. These include RNA aptamers, which can bind proteins or small molecules, and thus modulate cell physiology [1–3]. RNA aptamers are typically generated using in vitro selection methods, such as SELEX (Systematic Evolution of Ligands by EXponential Enrichment) [4–6]. However, only a small number of artificially derived, small-molecule-binding aptamers function robustly in mammalian cells and have been widely used for intracellular biosensors and other synthetic biology applications [7,8]. RNA aptamers that are developed by humans often lack the efficient folding that is seen with naturally occurring RNAs, which partially accounts for the discrepancy between good in vitro binding and poor intracellular function [9–11]. Naturally occurring RNA aptamers evolved over long time scales in a cellular context and were selected for their ability to be highly functional, which requires efficient folding. However, man-made aptamers are developed in a few weeks and only selected for their ability to bind a target ligand. As a result, they often show poor folding once introduced into cellular environment [12–14], which can therefore limit the progress of using aptamers in RNA synthetic biology applications.

Currently RNA aptamers are used primarily for in vitro applications. These include using aptamers as affinity tools, similar in many ways to antibodies [15–17]. Aptamers are also used in various types of analytical sensing devices, in which ligand binding induces an electrochemical or other signal [18,19].

However, aptamers are not widely expressed in cells to function in the cytosol to modulate cellular processes. Many publications have described examples of aptamer expression in mammalian cells, but these have not typically been replicated [20]. As a result, RNA aptamer-based therapeutics, which have been established for over 20 years, are lagging behind other RNA-based technologies such as small interfering RNAs, antisense oligonucleotides or guide RNAs despite aptamer’s unique ability to mimic antibodies and directly interact cellular proteins and metabolites [21,22]. The key problem is that functional RNA aptamers cannot be expressed at sufficiently high concentration. Recent studies have started to shed light on why aptamer expression in mammalian cells does not lead to effective modulation of cellular processes, and ways to fix this.

The first major problem is that RNA aptamers are vulnerable to exonuclease degradation. This problem has largely been overcome by expressing RNA aptamers in the circular RNA format [20]. Recent studies have directly assessed the exonuclease degradation problem and found that RNA aptamers have half-lives of 30–90 min and therefore only accumulate to low to mid nanomolar levels [23]. As a result, RNA aptamers cannot achieve the concentrations necessary to stoichiometrically bind and thus modulate target proteins. This problem has largely been overcome with a new expression system that allows aptamers to be expressed as circular RNAs [20], i.e. RNAs that have their 5’ and 3’ ends connected. Because they lack 5’ and 3’ ends, circular RNA aptamers are resistant to cellular exonucleases and can have half-lives exceeding three days [20]. This expression system, termed Tornado (twister-optimized RNA for durable overexpression), involves expressing the RNA aptamer with flanking ribozymes which then enable an endogenous cellular RNA ligase to catalyse 5’ and 3’ end joining [20]. This Tornado approach largely solves the nuclease degradation problem faced by many RNA-based technologies, and it has been adopted to enhance the function of fluorescent aptamers [11,20,24], the NF-κB-binding aptamer [20], ADAR-recruiting guide RNAs [25,26] and CRISPR guide RNAs [27].

Another major problem is that RNA aptamers fold poorly in the cellular milieu [23]. If only a fraction of the expressed RNA aptamer is folded, then the ultimate effect of the RNA aptamer will be reduced. The problem of RNA aptamer folding was discovered as a result of work with genetically encoded fluorogenic RNA aptamers [28]. These aptamers have to be folded in order to exhibit fluorescence inside cells. However, as described below, numerous lines of evidence have shown that the fluorogenic aptamers often show lower fluorescence than expected, and this was due to poor folding. Although in-cell RNA folding can be measured using complex assays such as SHAPE [29], the simplicity of measuring cellular fluorescence has allowed a wide variety of new approaches to be developed that can improve RNA folding in vitro and in vivo.

In this review, we discuss the problem of RNA aptamer folding and strategies for improving the cellular folding of RNAs. We first discuss new approaches for improving aptamer folding without the use of RNA scaffolds. These new approaches provide guidelines for future evolution of aptamers towards better cellular folding. We also discuss how RNA scaffolds can be used to improve folding of existing RNA aptamers. Inserting RNA aptamers into RNA scaffolds is a convenient way to enhance their performance in the cytoplasm.

Evidence for inefficient RNA aptamer folding

The idea that RNA aptamer folding is inefficient largely comes from studies of fluorogenic RNA aptamers, such as Spinach, Broccoli, Corn, Squash, Chili, Mango and others [11,28,30–34]. These RNA aptamers were evolved to bind otherwise non-fluorescent small molecule fluorophores resembling those normally found in GFP and related proteins [28]. When the RNA aptamer binds these fluorophores, the fluorophore becomes highly fluorescent, allowing the RNA-fluorophore complex to be detected using in vitro fluorescence assays or by fluorescence imaging in live cells [28]. We described the Spinach aptamer, which was the first RNA aptamer to be used to genetically encode fluorescence in a living cell [28]. The development of Spinach and the demonstration that aptamers can be used to encode fluorescence in cells led to the new and rapidly developing area of chemical biology focused on developing RNA-based imaging probes. For all fluorogenic aptamers, their ability to function in the cell is important, since they are specifically made for the purpose of cellular expression.

Fluorogenic aptamers (sometimes called fluorescent light-up aptamers, or FLAPs) are often made using in vitro techniques such as SELEX [4–6]. The SELEX technique focuses on finding library members that can bind a target molecule. The resulting human-made aptamers do not benefit from the millions of years of evolutionary pressure for efficient folding and function in cells that naturally occurring RNA aptamers experience.

Thus, perhaps not unsurprisingly, when fluorogenic aptamers were expressed in cells, they were less fluorescent than expected based on their in vitro fluorescence. Indeed, the initial realization about the importance of RNA folding for in-cell applications of aptamers came with studies of Spinach. Spinach produces fluorescence with high quantum yield in vitro, but lower fluorescence in cells [30]. As a result, Spinach was initially used to image RNA aggregates in mammalian cells, rather than single mRNA molecules.

In order to understand if Spinach folds efficiently, we developed RNA aptamer folding assays [30]. These assays only work with fluorogenic aptamers, but nevertheless are likely to provide insights into any SELEX-derived aptamer. In this assay, the absolute fluorescence of the Spinach-fluorophore complex is measured using a small, fixed amount of fluorophore, and excess Spinach. This way, even if most Spinach is unfolded or mis-folded into alternative structures, there should still be enough fully folded Spinach so that every fluorophore is bound by a Spinach aptamer. Thus, if 0.1 µM fluorophore is used and 10 µM Spinach is used, there should be 0.1 µM Spinach-fluorophore complex. In this way, the molar fluorescence of the Spinach-fluorophore complex can be measured and then used for a subsequent calculation for Spinach folding. This calculation measures the discrepancy between the observed fluorescence for a Spinach-fluorophore solution and the expected fluorescence if all the Spinach were folded. The calculation of Spinach folding is performed in this case by using excess fluorophore and a fixed amount of Spinach aptamer. For example, if 0.1 µM Spinach is used and excess fluorophore is used, only the fraction of the 0.1 µM Spinach that is folded will contribute to the overall fluorescence. The difference between this value and the expected fluorescence of 0.1 µM Spinach-fluorophore complex provides an exact measurement of the amount of Spinach that is folded.

These folding measurements revealed that Spinach was~32% folded in solution [30]. Importantly, this folding percentage can be highly influenced by buffer composition and temperature. Nevertheless, these experiments indicated that Spinach folding could be improved to further enhance fluorescence.

There were also hints that Spinach folding was impaired inside the cell. In general, the relative fluorescence of Spinach-expressing cells was low [28,30]. However, it was difficult to determine if this was simply due to low expression levels of Spinach since small RNAs are often highly unstable [23]. In principle, cells expressing an equivalent amount of GFP molecules and Spinach molecules could be compared for their fluorescence. However, fluorescent proteins are expressed at very high levels, whereas small RNAs exhibit very low expression levels [23]. It is difficult to obtain a direct one-to-one comparison based on expression levels. Thus, it was difficult to know if the low fluorescence of Spinach-expressing cells was a result of poor folding, or expression, or both. It should be noted that in addition to RNA mis-folding on its own, cellular RNA-binding proteins or helicases may also impair folding or even unwind the folded RNA. Nevertheless, the relatively low in vitro folding of Spinach suggested that folding should be addressed as one factor that could improve cellular fluorescence.

These observations led to a series of innovative approaches to improve in-cell aptamer folding, using the fluorescence of the fluorogenic aptamer as a simple and robust read out to infer the folding in cells. Many of these methods will likely apply to other types of aptamers, not just fluorogenic aptamers.

Rational mutagenesis for improving aptamer folding

The first approach for improving Spinach fluorescence involved rational mutagenesis. The initial Spinach aptamer showed ~32% folding [30] which could be improved to ~58% by making rational mutations in the Spinach sequence to remove bulges and reduce the presence of sequences that could potentially contribute to alternative folding structures [30] (Figure 1a). The improved Spinach aptamer, term Spinach2, also showed brighter fluorescence in cells [30] (Figure 1a). The increase in brightness can be measured by simply using the same exposure time during fluorescence imaging of cells expressing the initial Spinach aptamer versus Spinach2. After normalization for RNA expression levels, the increase in fluorescence of Spinach2 likely reflects improved folding. These data showed that increasing the folding in vitro can lead to improved function in cells.

Figure 1. Various approaches for maximizing the folding stability of the ultimate SELEX winner. (a) Using rational mutagenesis to improve aptamer folding. The initial Spinach aptamer was improved to Spinach2 aptamer by making rational mutations (red) in the Spinach sequence to remove bulges and reduce the presence of sequences that could potentially contribute to alternative folding structures. The improved Spinach2 aptamer showed brighter fluorescence in cells [30]. (b) Using low magnesium (Mg²⁺) concentration during SELEX and directed evolution in bacteria. Broccoli was selected using a few rounds of SELEX followed by a bacteria-based FACS screen of a partially selected SELEX library. Both the in vitro selection and the FACS sorting were performed in buffers with low magnesium (100 µM). The resulting aptamers showed much lower magnesium dependence than Spinach, and additionally showed increased intracellular fluorescence when expressed in mammalian cells [31]. (c) RNA-stabilizing ligands can act as molecular chaperones. the DFHBI fluorophore (light green) was modified so that it could achieve additional interactions with the RNA structure. The modified fluorophore is termed BI (dark green). The additional contacts further stabilize and trap Broccoli aptamer into the folded state, and thus push the equilibrium towards generating more fluorescent aptamer molecules. Broccoli-expressing cells cultured with BI showed more fluorescence than cells cultured with DFHBI-1T and other previously synthesized Broccoli fluorophores [24]. (d) Using highly folded natural riboswitch aptamers as scaffolds for SELEX libraries. The xpt-pbuX guanine (Gua) riboswitch aptamer, the Vc2 cyclic di-GMP (CDG) riboswitch aptamer, and the S. Mansoni hammerhead ribozyme were transformed into GR scaffold, CG scaffold and HR scaffold respectively [9]. The transformations were done by complete randomization of the junction region (red). Complete randomization means that each original nucleotide in the aptamer was mutated to either A, C, G or U with equal 25% probability. Complete randomization does not modify sequence length of the region. GR, CG and HR scaffolds were used to evolve aptamers that bind 5-hydroxyl-L-tryptophan (5HTP) and 3,4-dihydroxy-L-phenylalanine (L-DOPA) [9]. The add adenine riboswitch was transformed into add a scaffold by a novel randomization scheme called ‘the Sprouts-and-Clips’ (green) [11]. This method not only mutates the identity of each nucleotide, but also expands or contracts the sequence length, resulting in potential ligand binding pockets of different sizes. The new Squash aptamer evolved from the add a scaffold.

Other rational mutagenesis approaches took advantage of the structure of Spinach. The structure revealed the residues that were important for fluorophore binding as well as residues and base pairs that could be potentially swapped or removed since they were not essential for forming the fluorophore-binding pocket [35,36]. The mutations led to a smaller and more compact sequence which still retained full-length Spinach’s fluorescence, termed Baby Spinach [36].

It should be noted that the ease of detecting aptamer fluorescence in cells can vary considerably depending on the instrumentation that is used. Thus, it is not fully accurate to say that Spinach fluorescence is difficult to detect. The initial studies using Spinach and related aptamers were designed for the purpose of being compatible with very basic imaging equipment. The goal was to use imaging parameters that would be accessible to general users, rather than fluorescence microscopy experts who have access to more specialized and sensitive equipment. For this reason, it is difficult to compare Spinach imaging studies with studies in which RNAs are imaged using the MS2 system [37] where typically highly complex and highly sensitive instrumentation is used [38–40] to see fluorescence signals.

Directed evolution in low magnesium conditions

One hypothesis for the relatively poor folding of SELEX-generated aptamers in living cells is due to the relatively lower free magnesium concentration inside cells compared to the magnesium concentrations used in a standard SELEX experiment. Many SELEX experiments use 5–20 mM magnesium in the buffers [12,41–44], which can make the aptamers reliant on this high concentration of magnesium for folding. Notably, magnesium is a critical ion for RNA folding because it is very effective at neutralizing the negative charge of the phosphates in the backbone. RNA folding is highly limited by the repulsion between phosphate groups, which ultimately reduces the ability of different strands of RNA to come into proximity. Magnesium has a critical role in neutralizing this charge-charge repulsion [45,46].

However, the concentration of free intracellular magnesium is relatively low. Magnesium in mammalian cells is complexed with ATP, phosphate, and other negatively charged anions [47]. Therefore, even though reported concentrations of intracellular magnesium typically range between 0.1 and 1 mM [47,48], the available magnesium concentration may be lower. Thus, it is possible that selecting aptamers in buffers that contain lower magnesium may make RNA aptamers fold properly when placed inside mammalian cells.

We tested whether selecting aptamers in low magnesium would result in subsequent improvements in cellular performance, which led to the development of Broccoli, another green fluorogenic aptamer [31]. Broccoli was selected using a few rounds of SELEX followed by a bacteria-based FACS screen of a partially selected SELEX library [31] (Figure 1b). The FACS-based screen involved creating a library of the RNAs in the SELEX round, expressing one aptamer per bacterial cell, and then using FACS to identify cells expressing aptamers that showed high fluorescence. These bacterial cells were sorted in the presence of the DFHBI fluorophore, which becomes fluorescent upon binding Broccoli [31]. Both the in vitro selection and the FACS sorting were performed in buffers with low magnesium (100 µM) (Figure 1b). The resulting aptamers showed much lower magnesium dependence than Spinach, and additionally showed increased intracellular fluorescence when expressed in mammalian cells [31]. These experiments supported the idea that aptamers should be selected in low magnesium buffers in order to be prepared for the low magnesium environment of mammalian cells.

Importantly, these experiments comprise the first example of directed cell-based evolution of an RNA aptamer. By expressing libraries in bacteria and selecting for fluorescence, the best candidate aptamers could be rapidly selected. These candidates can then be mutagenized to create new libraries and selected for additional features with corresponding selection conditions. Since this approach uses FACS and relies on fluorescence, it is not immediately transferable to other non-fluorescent aptamers. However, methods that couple aptamer folding to a fluorescent readout could potentially enable this approach to be used for other types of aptamers. It should be noted that this selection was performed in bacterial cells since it is much simpler to achieve one library member per cell using bacteria. The bacterial cytosol is different from the mammalian cytosol and thus methods for directed evolution in mammalian cells for RNA would be highly desirable.

Notably, Broccoli has high sequence similarity to Spinach especially in the fluorophore-binding domain [49]. The reduced magnesium dependence likely reflects a reduced magnesium requirement for the RNA sequences outside of the fluorophore binding domain that provide overall structural support for the fluorophore-binding pocket [31].

RNA-stabilizing ligands as molecular chaperones

Another approach involves using the fluorophore ligand to enhance RNA folding. Importantly, RNA likely undergoes various transitions between folded, unfolded, and mis-folded states. The major challenge is to increase the thermodynamic stability of the folded state, making an aptamer less likely to adopt unfolded or mis-folded states which would simply reduce the overall fluorescence output. Each of the mutational approaches described above likely makes the folded state more thermodynamically stable.

Rather than further mutate the aptamer sequence, we reasoned that the DFHBI fluorophore could be modified so that it could achieve additional interactions with the RNA structure [24] (Figure 1c). Novel predicted Broccoli-binding fluorophores were discovered using an in silico screen of ~ 800 different DFHBI derivatives, each of which was docked into the fluorophore-binding site, as determined in the Spinach crystal structure. The virtual screen identified several DFHBI derivatives that were predicted to have enhanced binding. A subset of these molecules were synthesized, and one molecule, termed BI, led to markedly increased affinity for Broccoli. Additionally, Broccoli-expressing cells cultured with BI showed more fluorescence than cells cultured with DFHBI and other previously synthesized Broccoli fluorophores. Thus, BI enhanced the performance of Broccoli in living cells [24].

We found that BI promotes Broccoli folding by ‘rescuing’ partially folded intermediates. BI was able to bind a partially folded, but nonfluorescent Broccoli folding intermediate [24]. Upon binding, BI induced the fully folded conformation. In this way, BI rescued a portion of Broccoli that was not properly folded. Notably, this effect was only seen at 37°C, indicating that this partially folded form is likely due to partial thermal denaturation [24]. Other Broccoli ligands, such as DFHBI, were unable to rescue this partially folded Broccoli intermediate. The additional contacts made by the larger BI fluorophore likely account for its ability to ‘chaperone’ Broccoli to a folded state [24] (Figure 1c). Thus, in the unique circumstance when an aptamer binds a ligand, modification of the ligand could potentially trap folded states.

Because of the markedly improved Broccoli folding observed with BI, BI was used to image single mRNA molecules designed to contain multiple tandem repeats of Broccoli in their 3’ UTR [24]. The presence of multiple Broccoli aptamers allowed each individual mRNA to have an aggregate fluorescence that allowed individual mRNA molecules to be visualized in living cells using fluorescence microscopy [24]. These data indicate that simply including Broccoli aptamers in the 3’ UTR would not be sufficient for mRNA imaging since many of those aptamers would not be folded. The use of BI overcomes the folding problem to trap the aptamers in a folded state. As described below, the Broccoli aptamers in the mRNA also contained the F30 folding scaffold to further enhance the folding of Broccoli.

Using highly folded natural aptamers as templates for SELEX libraries

A final approach to evolve RNA aptamers that are well folded lead to the development of Squash [11]. Squash was not evolved from a completely randomized library, which is used for to create aptamers in most SELEX experiments. Instead, Squash was designed from the aptamer portion of an adenine riboswitch that is naturally found in bacteria (Figure 1d). Natural aptamers are particularly useful when considering RNA folding since considerable evidence supports the idea that many natural aptamers are highly efficiently folded. As an example, bacterial riboswitches fold in RNAs as the RNA is synthesized. The nascent RNA likely exists for only a few seconds before regulatory metabolites bind the riboswitch, inducing a conformation that can inhibit RNA polymerase in order to abort transcription in a timely manner [50].

Thus, it is desirable to exploit the highly efficient folding of natural aptamers. Batey and colleagues described the idea of using natural aptamers, but randomizing their ligand-binding pockets for SELEX [9] (Figure 1d). This concept was extended by us to take advantage of the highly efficient folding of the adenine aptamer in the development of Squash (Figure 1d). The ligand-binding domain of the adenine aptamer was randomized using a novel approach in which both the sequence composition and the length of the nucleotides were varied [9,11]. Importantly, the structural elements that allow the aptamer to achieve its unique conformation were preserved. Aptamers were then selected for their ability to bind to GFP-like fluorophores, resulting in Squash [11]. Notably, Squash showed high in vitro folding using the folding assay described above, as well as very high cellular fluorescence. These experiments supported the idea that using naturally occurring aptamers as scaffolds could ultimately lead to a better folding aptamer inside cells.

Importantly, the adenine aptamer normally functions in bacterial cells, but the evolved product, Squash, was used in mammalian cells. The crystal structure of the adenine aptamer shows relatively low magnesium dependence on folding, and the magnesium dependence was specifically tested and shown to be relatively low for Squash [11,51]. Thus, naturally occurring aptamers may be useful starting points for evolving new aptamers, but care should be taken to ensure that the aptamer is not reliant on high concentrations (e.g.>100 µM) of magnesium and can fold as efficiently in bacterial cells as in mammalian cells.

Overall, fluorogenic aptamers have been very useful for studying RNA aptamer folding and strategies to increase RNA folding in cells. Fluorogenic aptamers are useful for studying folding since only a folded aptamer can bind and induce the fluorescence of the small molecule fluorophore. Diverse strategies have been employed to optimize folding, resulting in general concepts that could be applied to other aptamers.

Using stable RNA scaffolds to promote RNA folding

RNA scaffolds are well-folded RNA sequences that have insertion sites into which an RNA aptamer can be inserted. We reasoned that RNA scaffolds could promote aptamer folding in cells. The idea is that the RNA scaffold will fold rapidly and efficiently, and thus force the 5’ and 3’ strands together to form a helical stem (Figure 2a). As a result, the folding of the RNA aptamer gets a ‘head start’ by forcing this stem to form, and preventing these strands from forming unwanted interactions with other parts of the RNA. Overall, an RNA scaffold limits the number of folding intermediates and increases the likelihood that the aptamer adopts the correct conformation.

Figure 2. Using RNA scaffolds to promote folding of artificial RNA aptamers. (a) Secondary structure of F30-Broccoli fusion. Broccoli was inserted into one arm of F30 by removing the terminal loop and replacing with Broccoli sequence [23]. (b) A novel design which uses the F30 RNA scaffold to develop sensors. In this example, S-adenosyl methionine (SAM) (dark blue) binding aptamer (light blue) was inserted into one arm of the F30 (orange) in such a way that the F30 arm was unstable and thus destabilized the entire F30 three-way junction region. The adjacent arm harbouring Broccoli aptamer (green) also became unfolded. Upon binding SAM, the SAM aptamer helical stem was stabilized, which promoted the folding of the entire junction region via Mg²⁺ ions and non-Watson-Crick interactions. Broccoli aptamer folds and binds DFHBI-1T to produce green fluorescence as a result [52].

It should be noted that the term RNA scaffold has also been used to describe RNAs that contain multiple aptamers to aggregate aptamer-binding proteins [53]. In this review, the term RNA scaffold (or RNA folding scaffold) is used to refer to RNAs that promote folding of aptamers.

RNA scaffolds such as GNRA tetraloops and U1A hairpins were initially used to facilitate crystallization of RNAs by increasing stability and reducing conformational heterogeneity [54–57]. Crystallization requires that each biomolecule be in the exact same conformation. Conformational heterogeneity would prevent the formation of crystals. As a result, flexible small RNA ribozymes and small RNA aptamers have been notoriously difficult to crystallize [56]. In contrast, much larger and more complex RNAs with a stably folded core, such as group I introns [54], have been more amenable to crystallization.

Subsequently, tRNAs were developed as RNA scaffolds [58]. tRNAs naturally fold into an L-shaped structure which is known to exhibit high thermodynamic stability. Importantly, tRNAs in cells typically contain diverse modifications; however, modification-free tRNA also is thermodynamically stable [58]. As a result, aptamers and other RNAs can be inserted into the anti-codon loop of tRNA and transcribed in vitro, and still show enhanced folding [58].

Although tRNA scaffolds are useful for in vitro transcribed RNAs, we found them to be unsuitable for expression in mammalian cells due to their ability to recruit tRNA processing enzymes [23]. The initial use of tRNA in mammalian cells involved tRNA-Spinach fusions [28]. Although this construct enabled genetically encoded fluorescence, the overall fluorescence was low. As described above, this was in part due to poor folding of the initial Spinach aptamer. However, in subsequent studies we showed that the tRNA-Spinach fusions are processed in cells by RNase P and RNase Z, which normally function to process endogenously transcribed tRNA precursors [59–61]. Since the tRNA scaffold is recognized as a tRNA, these enzymes cleave the RNA, thus facilitating its degradation. The cleavage of the tRNA also physically separates the tRNA from the aptamer, thus removing the scaffolding effect of the tRNA.

In order to replace the tRNA scaffold, we developed a new scaffold termed F30 (Figure 2a). F30 was developed from a three-way junction RNA from the phi29 bacteriophage RNA [62]. This sequence is highly thermodynamically stable and has been used extensively by Guo and colleagues in RNA nanotechnology to assemble RNA nanoparticles and RNA nanostructures [63,64]. Guo and colleagues demonstrated that aptamers inserted into phi29 exhibited increased folding [62]. However, phi29 is a bacterial sequence and only used for in vitro applications. It had not been used for mammalian expression. Inspection of the phi29 sequence showed that it contained a U-rich sequence resembling the termination sequence for Pol III, the polymerase that is commonly used for expressing small RNAs [23]. Expression of phi29 in mammalian cells showed that it indeed was expressed with a premature termination [23]. To overcome this, the U-rich sequence was mutated. The mutations were guided by the crystal structure of phi29 in order to ensure that the mutations would not affect critical structural interactions needed for folding or stability [23]. The resulting three-way junction sequence was termed F30. Fluorogenic aptamers fused into F30 showed markedly higher fluorescence in vitro and in cells compared to the non-scaffolded aptamers [23] (Figure 2a). Additionally, insertion of the aptamers into F30 prolonged their half-lives in mammalian cells to up to 80 min, compared to 30 min without the scaffold [23]. The increased stability may also reflect improved folding since folded sequences are presumably less accessible to intracellular nucleases. Overall, these studies revealed F30 as the first RNA folding scaffold compatible for use in mammalian cells.

Another scaffold, termed V5, has also been described to promote the folding of RNA aptamers in vitro [65]. This RNA is derived from the Vibrio proteolyticus 5S rRNA. However, when expressed in bacterial and mammalian cells, it showed evidence of intracellular cleavage, although to a lower degree than the tRNA scaffold [23]. Additionally, when compared with F30, V5 did not substantially stabilize aptamer folding in cells. For this reason, F30 has been the major RNA scaffold used in mammalian cells [20,24,66–68].

The F30 RNA scaffold can be used for sensor development

The F30 scaffold has also been used to develop new types of genetically encoded metabolite sensors [52] (Figure 2b). The development of Spinach and related fluorogenic aptamers has enabled the creation of genetically encoded ‘RNA sensors’ by us and others that comprise fusions of a biomolecule-binding aptamer to the fluorogenic aptamer [69–71]. The biomolecule can be either a metabolite or a protein. The initial designs used a simple transducer domain comprising a thermodynamically unstable helix connecting the fluorogenic aptamer to the biomolecule-binding aptamer [69]. The unstable transducer helix causes the fluorogenic aptamer to be unfolded. However, when the biomolecule-binding aptamer senses its cognate ligand via binding, its ligand-induced folding provides stability to fold the transducer helix, and then the fluorogenic aptamer. These designs were implemented in cells and were used to successfully image SAM dynamics [69], and subsequently the MS2 coat protein (MCP) in bacterial cells [70]. Notably, this concept was borrowed from the allosteric ribozymes developed by Soukup and Breaker that exhibited increased ribozyme activity after addition of various metabolites and small molecules [72]. By applying this concept to fluorogenic aptamers and then expressing these in cells, these studies revealed that aptamer folding can be regulated by small molecule ligands in live cells.

We recently described an alternative design which takes advantage of the intricate relationship between the arms of most three-way junctions [52] (Figure 2b). In these structures, the arms interact with each other via cations or via non-Watson-Crick interactions between the nucleotides. In this alternative sensor design, the biomolecule-binding aptamer was inserted in one arm of a three-way junction in such a way that the arm was unstable and thus destabilized the entire junction region. The adjacent arm contained Broccoli was therefore unfolded (Figure 2b). However, upon binding the metabolite, the aptamer helical stem is stabilized, which enables interactions with the other helical stem, thus stabilizing the folding of Broccoli (Figure 2b). This sensor design also works with other known three-way junctions, including three-way junctions Twort and H33 [52]. However, only the F30 sensor designed worked in living cells, while the other three-way junction sensor designs only worked in vitro. Overall, these experiments demonstrate a new use for the F30 scaffold in which the unique molecular interactions that confer efficient folding were regulated by metabolite-induced folding of an aptamer.

The folding methods described here were originally developed using fluorogenic aptamers, since improved folding of these aptamers is quantitatively and readily assessed by increases in fluorescence. However, these folding methods have been applied to other aptamers and RNA devices. These include newly developed fluorogenic aptamers [66], small-molecule-binding aptamers [10], protein-binding aptamers [20,67] and ribozymes [12,68]. As a result, both fluorogenic aptamers and non-fluorogenic aptamers benefit from using these methods. By increasing stability and minimizing alternative structures, both types of aptamers are able to function better in mammalian cells.

Future directions

A major hurdle for RNA synthetic biology is achieving well-folded RNA aptamers when expressed in mammalian cells. Emerging data demonstrate that aptamers are not well folded either in vitro or in cells. The folding in cells is particularly difficult to achieve because selection strategies typically rely on in vitro experiments that utilize buffers that cannot faithfully recapitulate the complex intracellular milieu. New experiments such as directed evolution in living cells [31,73,74], as well as the use of RNA scaffolds described here can help to achieve efficient RNA folding.

It is interesting to speculate on whether widely used RNA aptamers could function better if they are inserted into RNA scaffolds. One of the most commonly used RNA sequences in molecular biology is the MS2 hairpin [75]. The MS2 hairpin binds the MCP coat protein, and can therefore be used to recruit MCP-fusion proteins. However, it is unclear whether the MS2 sequence is efficiently folded in cells. In some studies, tandem repeats of the MS2 sequence were spaced out by linker sequences comprising 50 or more nucleotides in order to improve the folding of MS2 by preventing inter-aptamer hybridization [76]. Conceivably, RNA scaffolds like F30 could increase the folding of the MS2 hairpin and thereby increase the efficiency of MCP-fusion protein recruitment.

It is also interesting to speculate as to whether other RNA scaffolds besides F30 could be used to promote folding of aptamers. As described earlier, other three-way junctions have been used for the purpose of creating novel types of RNA-based sensors in which a metabolite-binding aptamer was inserted in one arm in order to regulate the folding of Broccoli in another arm [52]. Only sensors based on the F30 scaffold functioned in mammalian cells [52]. It is not clear if the other three-way junctions were unable to fold efficiently in a mammalian environment. Nevertheless, other three-way junctions could potentially be useful. It is worth noting that rigorous in-cell studies of RNA folding using methods such as SHAPE-Seq [29] have not been performed to determine the exact fraction of folded F30 or of aptamers in cells. These methods have the ability to reveal the existence of multiple conformational states based on correlated reactivites. Therefore, these types of experiments could eventually be useful to help benchmark the folding efficiencies of the three-way junctions as well as the aptamers that are attached to them.

In the future, directed evolution approaches in mammalian cells focused on optimizing RNA function, and presumably RNA folding, will be needed to design optimal RNA devices for mammalian RNA synthetic biology. Methods for continuous directed evolution of proteins, such as VEGAS [73], or related methods using adenovirus [74], could potentially be applied to evolve RNA and eventually to optimize RNA performance in cells.

Disclosure statement

S.R.J. is the co-founder and has equity in Chimerna Therapeutics and Lucerna Technologies. Lucerna has licensed technology related to Spinach and other RNA-fluorophore complexes.

Additional information

Funding

This work was supported by NIH grant R35NS111631 to S.R.J. Q.H. was supported by the NIH T32 GM115327 Chemistry-Biology Interface Training Grant to the Tri-Institutional PhD Program in Chemical Biology.