Engineered RNA Nanodesigns for Applications in RNA Nanotechnology

Abstract

Nucleic acids have emerged as an extremely promising platform for nanotechnological applications because of their unique biochemical properties and functions. RNA, in particular, is characterized by relatively high thermal stability, diverse structural flexibility, and its capacity to perform a variety of functions in nature. These properties make RNA a valuable platform for bio-nanotechnology, specifically RNA Nanotechnology, that can create de novo nanostructures with unique functionalities through the design, integration, and re-engineering of powerful mechanisms based on a variety of existing RNA structures and their fundamental biochemical properties. This review highlights the principles that underlie the rational design of RNA nanostructures, describes the main strategies used to construct self-assembling nanoparticles, and discusses the challenges and possibilities facing the application of RNA Nanotechnology in the future.

Introduction

The physical world presents a very different landscape at the nanometer scale. Physical phenomena such as inertia and gravity vanish from view; instead the interactions of matter and energy are dominated by other phenomena like wave-particle duality, uncertainty, and quantum electrodynamics. Nanotechnology is the engineering of functional systems at this scale, where the axioms that govern material and device design differ dramatically from those found at the macroscale level. More precisely, “nanotechnology” typically encompasses both the miniaturization of existing technologies to the nanoscale, and the discipline of engineering molecular-scale systems from the bottom up, producing constructs with fundamentally new qualities that revolutionize human engineering, from materials and manufacturing, electronics, and information technology, to medicine, biotechnology, energy, and even security. The array of goals and applications in nanotechnology intersects with an equally wide array of techniques and materials with their own emergent properties in the field. One such branch is concerned with the re-engineering of biological mechanisms, systems, and molecules in new forms with new utilities, broadly termed bio-nanotechnology.

The field of nanotechnology taking advantage of nucleic acids has its origins in the work of Nadrian Seeman and co-workers who have, over the past 30 years, spearheaded the development of DNA nano-object fabrication utilizing DNA self-assembly [1–6]. Briefly, DNA nanotechnology uses the nature of DNA complementarity for the construction of DNA tiles with discrete secondary structure by canonical Watson-Crick interactions (G-C and A-T base pairing), using a relatively small number of structural rules fundamentally based on Holliday junction motifs. The use of these rules has resulted in the engineering and characterization of numerous DNA 3D nanoscaffolds with different connectiviities [7–18] and the ability for some of them to promote targeted delivery by functioning as DNA nano-capsules [19–21] or DNA nano-carriers for other functionialities [22]. Another powerful technique called DNA “origami”, developed by Paul Rothemund [23], has been extended from its original scope for designing different 2D DNA shapes [24] and functional templates [25–28] to the creation of 3D objects such as a pyramidal tetrahedron [29], a functional nano-box [30,31] and a nanorobot [32], or shapes [33] relying on tensegrity [34] (structural integrity maintained by opposed tension in internal components) or helix bundles [35]. Similarly, Chad Mirkin and colleagues have done significant work using DNA oligonucleotides to form nanoparticle (NP) probes [36] by modifying colloidal nanoparticles with oligonucleotides which, upon the introduction of a complementary sequence, allow the self-assembly of nanoparticles into two- and three-dimensional architectures [37,38].

Presently, nearly all of the nucleic acid based polyhedral nano-scaffolds designed and tested in vitro employ DNA molecules as building blocks and have diameters greater than 15 nm [33,39,40]. Although these DNA nano-structures have demonstrated the potential to develop programmable scaffolds for nanotechnological applications [41], DNA biopolymers are not always able to mimic the diverse biological functions of its structural counterpart - RNA. RNA may also serve as an attractive biomaterial for a variety of nanotechnological applications, because of its unique structural, chemical, and biophysical properties which have some advantages compared to those of DNA [42,43]. Both DNA and RNA have intrinsic, manipulable features at the nanoscale and form systems with limited primary structures of four basic components. However, the presence of the 2’-OH group in RNA has a dramatic effect on its properties. Due to base stacking properties and the fixed C3’-endo sugar associated A-form helical structure, double stranded RNA offers improved thermal stability versus a DNA helix; this, combined with the ability of RNAs to form non-canonical base pairings, leads to the natural library of diverse structural motifs which in turn create a wide array of complex structures, many of which possess functional properties similar to proteins while retaining the simplicity of nucleotide primary structure [44–46]. A 20 base oligonucleotide has 10^12 possible sequences, in comparison to the approximately 7*10^26 possible sequences of 20 amino acids. This relative simplicity makes it easier to predict the function in synthetic macromolecule design [43]. Furthermore, RNA constructs have the potential to be integrated into native cellular machinery and thus, take advantage of expression within cells [47–49]. It should also be noted, however, that a diverse assortment of modified nucleotides, both natural and artificial, exist, and are particularly suited for integration with RNA structures, and present their own further challenges and opportunities. Natural modified nucleotides, such as pseudouridines, may simultaneously provide another tool for designers while also increasing a system’s complexity and thus the difficulty of application. Similarly, synthetic nucleic acids with promising properties, for example locked nucleic acids (LNAs), expand the library of possible base components and possible functions in nucleic acid nanodesign.

RNA was first characterized as a messenger in transcription in the 1940s and 50s [50–52] and its first function described in the 1960s [53,54]. It was twenty years later that the work of Tom Cech and Sidney Altman demonstrated the self-splicing of a ribosomal RNA precursor, and revolutionized the study of RNA [55,56]. Where before catalytic potential was thought reserved exclusively for proteins, this discovery immediately initiated a cascade of other multitasking functions assigned to RNA. The later discovery by Andrew Fire and Craig Mello, regarding the ability of RNA to selectively control gene expression in animals through the process of RNA interference (RNAi) [57], has cemented the place of RNA as an extremely important molecule for study with far reaching applications in medicine, science, and engineering. In particular, the capacity to understand and create customizable, functional RNAs has led to the genesis of a subfield of nucliec acid nanotechnology - RNA nanotechnology, pioneered by Neocles Leontis, Luc Jaeger, Eric Westhof, and Peixuan Guo - aiming to tackle obstacles in nanomedicine and synthetic biology [42,58,59]. The potential for medical applications of RNA nanotechnology, is exemplified by currently over 20 different RNA-based therapeutics in clinical testing [60]. The majority of these are now in phase I or II clinical trials, with none yet reaching phase III. It is important to note that the first efforts in applying the breakthroughs in RNA research discussed above were met with middling success that was far outweighed by the hype that surrounded it, however, the application of RNA has still progressed extremely rapidly [61]. This fact, taken together with the burgeoning of multiple startup companies makes the development of novel RNA therapeutics very real. RNA nanotechnology is uniquely able to harness the variety of catalytic functions RNA performs as well as organize these functions with a precise stoichiometry and orientation in 3D space.

Compared to proteins and some synthetic materials, RNA nanomedicine offers the potential for high specificity and low toxicity treatments of many diseases through the utilization of naturally occurring mechanisms [58,60,62] such as RNAi. Synthetically designed RNAs have the potential to form dynamic regulatory platforms for metabolic pathways and other natural systems [63]. They also can specifically recognize folded RNAs [64] and small molecules [65], regulate pathway expression [66,67], and contribute to programmable genetic circuitry [68–70].

The basis of RNA nanotechnology

The breadth of functionality in biologically occurring RNAs provides a very powerful foundation for the bottom up design of nanodevices in the form of various structural motifs and functional elements which operate in native biological systems, and can be derived from X-ray and NMR structures of natural molecules [59,71–78]. This repertoire of natural RNA motifs is augmented by new motifs with novel binding selectivities or enzymatic capabilities, obtained by in vitro selection, also called Systematic Evolution of Ligands by Exponential Enrichment (SELEX) [79,80], in which randomized pools of oligonucleiotides are iteratively selected, amplified, and mutated to enrich high-affinity binding to targets; as well as through rational design from biochemical principles [74,77,81]. Finally, all of these motifs can then be categorized into publicly available libraries for use in RNA nanotechnology.

The foundations of RNA nanobiology draw from several areas (see Figure 1) related to the understanding of RNA structure and function that reinforce each other and together can significantly speed-up the design process for determining the sequences needed for the correct folding and assembly of RNA nano-objects. In many cases the RNA design process essentially works in reverse of standard RNA structure/function determination practices. Normally RNA sequences are provided and the issue at hand is to determine the RNA secondary structure and ultimately the three-dimensional structure of the RNA. This is frequently accomplished by X-ray crystallography or nuclear magnetic resonance, though RNA secondary structural information, whether derived from experimental techniques or computational algorithms, is extremely helpful in determining the 3D structure of an RNA. Once characterized, the three-dimensional structure then provides a platform from which function can be determined and understood. Conversely, RNA nanodesign first decides upon the three-dimensional shape and position of scaffold structures and functional elements and only then is the secondary structure (or lack thereof), and ultimately primary structure determined. For many years efforts have been focused on the development of algorithms for the computational prediction of RNA secondary and three-dimension structure, the latter being done either de novo and/or with input from secondary structures. Some of the characteristics of RNA three-dimensional structures have also been studied experimentally with for example single molecule imaging or computationally with molecular dynamics. These methods have both contributed to and have utilized information obtained on structure and function of viruses and non-coding RNAs (e.g. the ribosomes). Thus, the knowledge obtained from these extensive studies on RNA structure and function is now being applied to the design of RNA-based nanoparticles. Figure 1 depicts the required knowledge flow for the design and ultimately the experimental assembly of RNA nanoparticles. Note that the arrows flow in both directions since feedback in all the areas depicted enhances each of the domains themselves.

Figure 1.

Components of modern RNA nanotechnology.

For the purposes of design, all modular components of RNA nanotechnology can be divided into two semantic classes: functional and tectonic units (Figure 2). However, these classes are not necessarily mutually exclusive. The first class represents RNA based functionalities with desirable chemical or biological activity including ribozymes [82,83], aptamers [84–86], riboswitches [87,88], and RNAi inducing agents (e.g. short interfering RNA (siRNA), micro RNA (miRNA), etc) [89,92] as well as functionalities with a non-RNA nature (proteins, small targeting molecules, gold NP, etc) [93–96], discussed in detail in Functionalization and Modification, below. The second class encompasses tectonic units that can be split into structural motifs and interacting motifs possessing discrete secondary and tertiary structures which can be combined to rationally produce self-assembling nanoparticles with predictable, multi-dimensional structures in vivo or in vitro. Some examples of structural motifs are double-stranded RNA helices, C-loops [72] (which increase the helical twists), kink-turns [97], right angle motifs [78], tRNA motifs [98] (the latter three bends within an oligonucleotide strand); as well as three [76,99], four [100,101], and five way junctions [100]. Examples of interacting motifs are tetraloop-receptors [74] (an interaction between a four-base hairpin “tetra-loop” motif and a “receptor” structured internal loop motif), kissing loops [102,103] (coaxial stacking motifs), sticky ends [104,105], and paranemic motifs [81] (an interaction of “crosshatched” stacking helices). Nanoscaffolds assembled from tectonic unit monomers can be readily ornamented with at least as many independent functional units as the number of tectonic units from which they are composed (see Functionalization and Modification). The functional units could be multiple copies of the same compound for increased potency, a collection of different therapeutic agents in a manner similar to cocktail therapy, signal or regulatory elements, or any combination of the above, dependent only on the objectives of the designer [106,107]. This combination of specific, programmable supramolecular structures with diverse functional units in a single, entirely RNA compound, allows precise positioning of different functional modules in three-dimensional space. This, for example, can provide an avenue to address the vital problems of stability, targeting, and multivalency in nanomedicine [95,102,103,107], and can form the basis of preprogrammed arrays [43], mediate the growth and distribution of other NPs [94,108], or form patterned scaffolds for tissue engineering or other applications [47].

Figure 2.

Schematic representation of modular components of RNA nanotechnology with some examples of different units and resulting nano-scaffolds and functional RNA nanoparticles.

While other NPs capable of multivalency, degradation resistance, and specific targeting have been developed [109] (see Functionalization and Modification), RNA nanotechnology offers numerous advantages for nanomedicine. It has greatly enhanced simplicity and biocompatibility when compared with peptide, liposomal, or synthetic nanomaterial conjugation for RNA-based functionalities; RNA nanoparticles can be equipped with custom tailored combinations of therapeutic agents or biosensors, which show promise to improve targeting and penetration of cells or tissues safely and efficiently [76,95,107], provide “cocktail”-like multi-target therapies [106], advanced imaging or imaging concurrent with therapeutics [110]. In addition, they can do so while guaranteeing strict control of the 3D orientation and stoichiometry of those agents [102,104]. Finally, therapies based on these principles rely on mechanisms distinct from existing drugs and thus may simultaneously improve the therapeutic ability of existing treatments and create new avenues for the treatment of drug-resistant or otherwise “untreatable” diseases [60,111,112].

Fundamentally, RNA nanotechnology can create customized nanostructures and devices de novo, with suitable properties for scalable production and myriad potential applications through the design, integration, and reengineering of powerful mechanisms that rely on a variety of tectonic units, diverse functional units, and underlying nucleotide structural principles [106]. RNA nanoscaffolds can be built to connect and coordinate multiple functionalities, including functional RNAs, and so operate as a robust, predictable platform for RNA-based therapeutics and synthetic biology. Thus, the ability to realize the promise of RNA Nanotechnology depends on the strategies used to rationally construct artificial RNA architectures with controlled folding and synthesis of multifunctional components inside and outside cellular systems.

Current strategies for nanoparticles design

There are two main strategies (Figure 3) used to engineer the current generation of programmable, robust 3D RNA nanoscaffolds, both of which are capable of generating 3D RNA structures that have been functionalized with proteins, siRNAs, and RNA aptamers.

Figure 3.

RNA architectonics and single-stranded tiles approaches.

RNA architectonic approach (tectoRNA).

The first is a bottom-up approach in which RNA molecules are reduced to minimal, modular units possessing discrete secondary and tertiary architectures [59,113–115]. Large libraries of these tectonic units (or tectoRNAs) are used to design self-assembling supramolecular structures based off of tertiary interactions of individual modules. This process, also known as RNA architectonics, is able to create architectures of desired sizes and shapes with a remarkable degree of structural control and has no analogous approaches in DNA nanotechnology [59].

In this strategy (Figure 3a), ideal 3D nanoscale structures are designed through “mosaic style” positioning in silica of tectoRNAs necessary to achieve the desired shape. This approach has been mostly manual, wherein a researcher or designer will typically make an executive decision based off of available or known tectoRNAs. However, recently several computational methods have been applied to aid in the design (see Computational tools and methods). Tectonic units are a broad class possessing different structural and interactive properties, and can be further categorized into subclasses with diverse variations and mutations; there are limitless possible combinations of tectoRNAs, and therefore possible architectures. That said, certain fixed constraints within a given sequence are necessary to ensure that 3D hydrogen bonding intrinsic to the tertiary structure is uninterrupted, and may limit the potential modifications of any given nanostructure. A secondary structure is then extrapolated from the 3D model and used to rationally design the set of primary sequences for the monomers involved in the formation of the particle. These structures can be characterized through biochemical experimentation and visualization techniques for the purposes of tuning designs or investigating structural systems with limited initial data.

Modular RNA units forming functional dimers [64,116,117], trimers [ 95], hexamers [73,102,103,106,118], and octameric antiprisms [93], all of which have been functionalized, have been designed in previous work using the RNA architectonics approach. Further work has produced non-functionalized nano-squares [98,105] as well as 1D and 2D arrays consisting of filaments [101] and squares [43], respectively.

RNAs that naturally assemble into macromolecules can be re-engineered for RNA nanostructure design. An illustrative example of this is the packaging RNA (pRNA) of the phi29 bacteriophage packaging motor [119,120], which can be assembled into different multimeric units via loop-loop interactions, and have demonstrated the capacity for multivalent functionalization [42,121]. Similarly, naturally occurred kissing loops [73,102,103,106] or tRNA motifs [93] have been utilized extensively in functional RNA nanodesign. Another example, the non-coding DsrA RNA from E. coli contains a native palindrome sequence which assembles into a regular extended structure reminiscent of extended lattices in DNA nanotechnology [122]. Our knowledge and understanding of RNA is expanding continuously, and ongoing research will continue to reveal other structures which will feed back into the motif libraries at our disposal (see Computational tools and methods).

In addition to the efficiency of using pre-existing motifs, the RNA architectonics strategy has numerous advantages in structural control and stability. However, oligonucleotides used in architectonics are often quite long and may require complex folding processes which can inhibit production, and are, broadly speaking, limited to the shapes derivable from available motifs and assembly conditions. Thus, an alternate strategy is to construct nanoparticles from shorter sequences more amenable to chemical synthesis and modification.

Single-stranded tiles approach.

The second strategy (Figure 3b) uses relatively short (26–52 nt) single-stranded RNAs, which are designed to avoid forming stable secondary structures and assemble without any tertiary interactions via thermal annealing [102,104,106,123–125]. This process may in some ways be qualitatively similar to the approach used to create nanostructures in DNA nanotechnology [5]; however, due to the higher thermodynamic stability of an A-form helix and increased stability of RNA non-canonical base pairs, RNA molecules can be trapped in alternative conformations much more easily then DNAs. This brings additional challenges to the successful design and formation of RNA 3D nano-objects [104] as it requires a deeper understanding of the mechanisms that dictate RNA folding and kinetics. On the other hand, RNA complexes demonstrate significantly higher thermal stabilities while having a lower chemical stability than their DNA counterpart. Moreover, this approach allows one to swap the RNA strands with corresponding DNA analogs and by altering the ratio of RNA to DNA strands entering the composition of nano-scaffolds one can easily fine-tune their stabilities to the necessary level [104].

The technique relies on a computational approach wherein RNA sequences are optimized in silico to maximize the fitness of the theoretical sequences with their desired final RNA structures while following designated sequence design constraints. This has the effect of distributing sequence constraints of the NP throughout the system, in contrast with NPs constructed from tectonic units, where dependencies are typically concentrated in junctions and loops interspersed with “open” helices [104,123]. An example of this methodology are the RNA nanocubes [104]; 3D models were generated using Nanotiler (see Computational tools and methods), and the sequence optimized via randomization and Monte Carlo optimization algorithms.

Designs of single-strand assemblies have been successfully used to generate trimers [125], tetramers [107,123], hexamers [104,106], octamers [105], decamers [104] that can be functionalized with aptamers, siRNAs, or non nucleic acids based functionalities.

Computational tools and methods

The construction and characterization of RNA nanostructures to date have relied heavily on the manual assembly and tuning of RNA units into nanoparticles based on preconceived designs. This approach, however, has a significant cost in both time and money even with the assistance of 3D computer visualization technologies. In response to limiting costs as well as an increasing demand for more complex designs, a variety of computational approaches have been developed to provide a more effective route for robust RNA NP design. This approach and the fostering of synergy between computation and experiments, has been pioneered by Bruce Shapiro. As discussed above, RNA nanoparticles can be designed using architectonics and single-stranded tile approaches, and while these strategies are complementary, each is based off of unique principles which demand different computational methods.

RNA Motif Databases.

RNA architectonics are dependent on the characterization and categorization of tectonic units from which to model and construct a nanoparticle and thus, the organization of tectonic units into searchable databases is a powerful tool for architectonic design [59,124,126]. One type of database necessary for RNA nanostructure design categorizes the sequence parameters of tectonic units: any given tectonic unit contains specific sequences necessary for it to retain its desired shape and function, thus, obligatory sequence parameters largely determine a tectonic unit’s capacity for structural manipulation [59]; other databases classify RNA structures by function, tertiary interactions, or 3D geometry, as in the Structural Classification of RNA (SCOR) and nucleic acid database (NDB) [127–129]. Similarly, the RNAJunction database categorizes tectonic units on the basis of documented geometries and recurring base- and tertiary- interactions not limited to canonical base pairing. RNAJunction currently consists of over 13,000 junction motifs and kissing loops derived from the Protein Data Bank (PDB) that are defined from 2-way (bulge and internal loops) up to and including 9-way junctions. A notable feature of RNAJunction is its ability to search for motifs such as kissing-loop interactions and junctions with specified angle ranges between emanating helical stubs by which other helices can connect.

3D Modeling tools.

In addition to databases, there are numerous design tools that help arrange, join, and model RNA structural elements in a 3D workspace [124,126,130–132]. There are numerous general purpose molecular modeling packages, such as Chimera [133], pyMol [134], and the Accelrys (http://accelrys.com/) suite. Each of these programs provides its own unique capabilities which may have uses for RNA nanodesign, however, they are typically incapable of specifically addressing the demands of RNA nanodesign. As an illustrative example of this, Chimera is capable of altering the geometry of a visualized nanoparticle by rotating a base pair on its axis, however, it does so unphysically; it does not “understand” RNA mechanics, and so the design must then be imported into a different program before any nanoparticle designed using this method could form. Alternatively, tools specifically designed for use with RNA and in nanodesign exist, which can simplify this process. One such tool is NanoTiler, which can generate defined nanoparticles from tectonic units found in the RNAJunction database or elsewhere These units can be connected by “standard” A form RNA helices of varying lengths [124,126]. Nanotiler allows the topological specification of desired structures and the sequence constraints necessary to conserve desirable tertiary interactions. It can also create do novo structures to fit design goals, and apply translation, bending, and twisting deformations to maximize “goodness of fit” in accordance with natural RNA dynamics. Moreover, in conjunction with a sequence generator, the program has the capacity to create sequence sets derived from the initial 3D atomic models which have demonstrated a high probability to self-assemble into desired constructs under experimental conditions. NanoTiler was used to design the nanocube and nanoring shown as examples in Figures 1 and 2, and the cube corner elements are an example of a motif created de novo when no natural motif with the desired property could be found. The structures generated by NanoTiler can then be imported into an associated sequence optimization program which can generate sequences that were used in the self-assembly experiments.

Another modeling program is RNA2D3D [131]; while NanoTiler represents a class of tools for the construction of de novo structures, RNA2D3D allows for rapid 3D visualization of supramolecular structures from primary and secondary structure information, and provides tools to subsequently adjust the structure, eliminate steric clashes, and optimize its energy and dynamics. Given the secondary structure pairing information and the primary sequence, the program transforms the paired regions of a secondary structure into idealized helices, and then integrates helical geometry with adjacent loops, automatically stacks user-selected helices, and facilitates modeling of pseudoknots and arbitrary tertiary interactions. RNA2D3D’s capacity to incorporate fragments of experimentally determined structures into the full model, connect multiple RNA chains via kissing-loop interactions, and dynamically add base-pairs to the selected helical regions are intrinsically useful in nanostructure modeling applications.

Production and assembly

Once a functional RNA nanoparticle has been designed using either strategy, it can be easily produced in vitro through RNA’s native self-assembling properties; however, as each strategy relies on different assembly principles their respective approaches to nanoparticle production present different challenges. RNA molecules for assembly can be purchased directly, however current technologies limit the length of RNAs available through commercial synthesis to strands shorter than approximately 80 nucleotides. This is a particular challenge for nanoparticles designed via RNA architectonics because they rely on predefined motifs which are often greater than 100 nts in length, which may be further complicated for nanoparticles designed using either strategy, as the functionalization of a nanoparticle typically involves the addition of relevant sequences to individual strands. An alternative approach is to transcribe RNA strands using T7 RNA polymerase with purchased PCR-amplified DNA templates, as DNA templates can be produced much more cheaply and in longer strands than RNA [106]. RNA products may then be retrieved from a transcription mixture via gel purification and stored.

Currently, there are three major approaches allowing the one-pot production of functionalized RNA nanoparticles depicted in Figure 4: RNA architectonics, single–stranded tiles and co-transcriptional assemblies.

Figure 4.

Schematics showing three major routes of functional RNA nanoparticle production: RNA architectonics, RNA single stranded tiles and RNA co-transcriptional assemblies.

In the case of RNA architechtonics assembly (Figure 4, upper panel), the use of natural RNA tertiary motifs normally requires pre-folding of all individual monomers before their assembly. In this approach, heating the mixture of strands at 95 °C for several minutes causes the melting of all hydrogen bonds thus eliminating undesirable RNA interactions. The heating step is normally followed by the immediate step of placing samples on ice (snap cool) thus guaranteeing the formation of the most thermodynamically favored helices and secondary structures of RNA motifs. Addition of the assembly buffer, containing divalent magnesium ions, promotes stabilization of tertiary RNA motifs and the further formation of intermolecular hydrogen bonds between corresponding regions on different strands of RNA at slightly elevated temperatures (~30°C).

In contrast to RNA architectonics, constructs designed intrinsically for a single-stranded tiles approach (Figure 4, middle panel) disallow secondary and tertiary formation in monomeric strands. This approach, based only on canonical Watson-Crick interactions, utilizes relatively short (26-48 nt) single-stranded RNAs as the nanoscaffold strands. The assembly of these nanoscaffolds normally requires heating the mixture of strands at 95 °C for several minutes (similar to architectonics assembly) followed by the immediate snap cooling to the temperature slightly below the melting temperature of the nanoscaffolds (e.g. snap cool to 45 °C in the case of the nanocubes) and addition of assembly buffer. This protocol guarantees the formation of the most thermodynamically favored structures eliminating all possible undesirable interactions.

Recently a generalized methodology for the production of functional RNA nanoparticles during in vitro transcription with T7 RNA polymerase (Figure 4, lower panel) has been developed [77,102]. This approach eliminates multiple steps currently associated with RNA NP production: individual transcription of each template encoding the functionalized NP subunits by bacteriophage T7 RNA polymerase, purification of the resulting RNAs, combining all the RNA components of the NP in equimolar quantities, their thermal denaturation, and refolding of the NP. Notably, the denaturation and refolding conditions depend on the RNA sequence type of the design approach (as described above), and, therefore, require individual optimization for each RNA NP. The complexity of the currently existing NP production protocols impedes their streamlined mass scale production. An alternative, less complex and more universal approach for the production of RNA NP has recently emerged with the spontaneous formation of the functional NP (nano-cubes and nano-rings) during transcription of a mixture of DNA templates encoding all the individual RNA components. Such co-transcriptional assembly of the NP bypasses the laborious and time-consuming steps of individual RNA strand synthesis and purification with further heat denaturation and renaturation. However, the detailed molecular mechanism of co-transcritonal folding of the RNA NP and the factors affecting efficiency of this process remain to be determined.

Functionalization and modification

The scaffolds generated using these complementary design strategies have the capacity to be easily tuned for functionalization in new applications for nanomedicine and synthetic biology because of RNA’s characteristic structure, programmability, controllable chemical and thermal stabilities, sequence flexibility, rich tertiary structure, and self-assembly [106]. While some of these applications, such as patterned superstructures for metallic nanoparticle distribution [108] and scaffolds for nanocrystal formation or tissue engineering, don’t necessitate biologically or chemically active components, one of the most valuable properies of RNA nanoparticles is precise, programmable, and addressable integration of important functional RNAs. This value arises from the issues of stability, targeting, and multivalency discussed above: nanoscaffolds show increased controllable programmability and addressability allowing a designer to precisely control the stoichiometry and position of small molecules within a particle as well as integrating multiple functionalities into one unit without compromising those functionalities.

Further, nanoparticles may require additional modifications to the underlying structure based on the conditions of the application for which they are being designed, to improve yields, stability, kinetic favorability, as well as allowing the hybridization of functional elements or scaffolds with other nanostructures or devices.

Some examples of RNA nanoscaffold functionalization.

The conjugation of nanoparticles with RNAi inducing agents has seen a substantial increase in interest in the last several years [42,60,62,121,135]. Briefly, RNA interference, or RNAi, is a process by which the expression of a gene can be silenced through the introduction of a double-stranded RNA (dsRNA). This dsRNA can be processed by the endonuclease Dicer into a short (21-23 nt) duplex known as short interfering RNAs (siRNAs), which is then loaded into the RNA induced silencing complex (RISC). Half of the duplex, the antisense or guide strand, is used by RISC to selectively cleave a target mRNA, while the sense or passenger strand will be discarded [57]. With the proper choice of a target, a short RNA can be designed to post-transcriptionally silence the expression of any gene. There are a variety of different dsRNAs that can activate RISC each of which have their own properties [60,136]. The most commonly used are 21-mer siRNAs. However, the functionalization of RNAi-inducing RNA-based NP can be done using elongated siRNA duplexes, so-called “dicer substrates” [137]. Using dicer substrates will allow the siRNAs to be released inside the cell through the process of dicing. Multiple siRNAs for combinatorial RNA interference [138,139] can be co-delivered together using the RNA NP approach [106]. RNAi has its own set of challenges to implementation which are discussed extensively elsewhere [62]. An example of this is RNAi saturation, in which induction of RNAi with introduced agents can up-regulate the RNAi for hundreds of non-targeted genes; this is distinct from the broader challenges of stability, biodistribution, toxicity, etc, which effect RNA nanomedicine and nanomedicine in general, but is conversely representative of the way in which any specific nanodesign which utilizes existing mechanisms must also address the challenges facing that system as well. Nonetheless, functionalized nanoparticles containing siRNAs have been demonstrated to possess significant silencing potential and may offer solutions to the challenges of specific targeting [139].

Ribozymes, which possess catalytic properties, are capable of cleaving RNA substrates and thus regulating gene expression, targeting viral RNA, synthesizing RNAs, and operating in biological logic circuits [140]. A ribozyme functionalized RNA NP could be constructed for therapeutic potential. However, ribozymes could also be employed to regulate the behavior of the nanoparticle itself, or other devices in a circuit or system; for example a self-cleaving ribozyme coupled with an aptamer could conditionally activate a nanoparticle through allosteric induction [141]. This latter function is of particular note as the RNA nanoparticle’s characteristic capacity for multivalency readily recommends it for the engineering of complex behaviors. Such a “smart” particle could overcome challenges of localization, concentration, and other limitations where independently acting functional units are non-ideal or prohibitive.

In a related approach with promise for engineering “smart” nanodevices, nanoparticles can be functionalized with aptamers for high-efficiency cell sorting, and to achieve imaging or therapeutic delivery to targets flexibly and specifically [85]. Aptamers are oligonucleotides with well-defined tertiary structure that bind specific ligands, which can be used to target disease relevant markers [142–144]. Randomized RNA pools are screened via Systematic Evolution of Ligands Exponential enrichment (SELEX) to select aptamers for a given target. Aptamers have been concatenated with RNA NP for targeted cancer cells in preliminary studies [20] and tracking of RNA NP assembly [104], as well as hybridized with siRNA [142] for targeted silencing in cancer cells.

Targeted delivery and conditional activation of therapy have the potential to dramatically mitigate off target effects in nanomedicine, which can be a serious issue with numerous types of therapies, especially for nanoparticles administered systemically. Off-target siRNAs can down-regulate important functions in non-disease targets, as well as cause RNAi saturation. Current research suggests that pooling siRNAs together, as in an RNA nanoscaffold reduces off target effects associated with individual siRNAs due to the possibility of lowered concentrations upon delivery and simultaneously increases potency.

Riboswitches are a class of RNAs related to aptamers, which can bind small molecules and regulate the behavior of gene expression, and can be re-engineered to perform a variety of other tasks including detecting toxins, controlling the behavior of bacteria, and operating as Boolean logic gates. Like aptamers and ribozymes, nanodevices could be functionalized with riboswitches as control mechanisms to modulate their own or other behavior in vivo [88,145].

Nanoparticles can be functionalized by the integration of functional unit sequences into the backbones of particles in cases such as the nanoring, where structural motifs are interspersed with open helices [102,103,106]. However, more commonly a nanoparticle is designed with 3’ or 5’ toeholds to which functional units or their scaffolds can be attached; this approach increases the modularity of the design as well as opening up possibilities for further modifications depending on the particular goals of the design.

Chemical modifications and variations

Additional modifications to the chemical structure of a nanoparticle are used to ensure proper folding, increase thermal stability, provide resistance to degradation, and, as with functionalization, broaden the field of potential applications of the technology through the variation and combination of modifications. RNA is an inherently fragile molecule; though it’s increased thermal stability is ideal for consistent, effective, secondary and tertiary structure formation, much of its structural and functional diversity is due to its chemical instability. The active properties of introns, ribozymes, and riboswitches, for example, are dependent on cleavage or allosteric response. Though RNA is resistant to certain conditions such as degradation in endosomal cytosis, it is a chief target of the immune system, can be easily degraded by RNAases present in sera, and rapidly undergoes renal filtration due to its negatively charged backbone [146]. Unmodified siRNAs, and other functional RNAs, drop below detectable concentrations within minutes when dosed systemically. The ability to administer the treatment systemically is important for many therapies, and further, any forward-looking application in synthetic biology or nanomedicine that proposes to operate in vivo must address the issue of stability, as well as those of quality control, visualization and characterization [42,102,106].

Different strategies have been employed to increase the chemical stability of RNA. The use of nanoparticles is itself already a dramatic step to this end. Folding and assembly of tertiary structures provides kinetic and thermal stability by definition. The assembly of the individual RNA subunits through kissing-loop interactions in the nanoring provides increased ribonuclease resistance not afforded to the individual unassembled subunits [58,103], for example, and the difference in particle size provides some protection from renal filtration, which primarily affects molecules less than 10nm across [109].

Circularization of a nanoparticle, or the sequestering of 3’- and 5’-ends within the larger structure, is another option that provides protection against ribouncleases [103,147], but by far the most applied strategy is the modification of the 2’ hydroxyl group of the ribose of individual nucleotides in the RNA strand [60]. This modification provides powerful protection against phosphate backbone cleavage, which has demonstrated retention of self-assembly and function versus unmodified RNA, in the case of modified pRNA, and the latter in siRNAs [102,106,125]. 2’-fluorinated dUTPs can be incorporated into the backbone of an RNA strand during transcription by mutant bacteriophage T7 RNA polymerase with reduced yields, or, with wild-type T7 polymerase in the presence of Mn²⁺ for comparable yields to unmodified RNA transcription. siRNA functionalized nanoparticles produced and assembled co-transcriptionally using this method have demonstrated increased resistance to nucleases in blood sera while maintaining their silencing ability [102].

Distinct from alterations to the RNA of a nanoparticle, a nucleic acid scaffold or component may be hybridized with other nanomaterials to form complexes with distinct characteristics in comparison to either component molecule. A new strategy for nucleic acid nanotechnology in this vein is the self-recognizing RNA-DNA hybrids (Figure 5) [148]. These hybrids contain complementary ssDNA toeholds and are specifically designed to carry multiple split functionalities (e.g. RNAi inducers, FRET pair of dyes, aptamers, etc) which can be easily activated inside target cells upon hybrid re-association with ssDNA toehold recognition. This novel approach can greatly benefit and bridge together the emerging fields of RNA and DNA nanotechnologies and has multiple advantages such as additional control over the functionalities. Further, R/DNA hybrids have been shown to have very high resistance to degradation in sera [149], and research has shown that RNA-DNA NPs would have additional advantages; DNA molecules are more chemically stable and therefore, cheaper to synthesize and more amenable to chemical modification than RNA. Thus, any modifications can be made to the DNA strand without interfering with RNA-DNA assembly, or disrupting the structure and function of released RNA-based functionalities [148]. Such a system could allow for conditional release or activation of functional units, improved targeting specificity, improved visualization in vivo, and increase overall quantity of functional units or functionalities, all while preserving the integrity of the individual subunits.

Figure 5.

Schematics explaining the simultaneous activation of multiple split-functionalities (RNAi and FRET) by re-association of two cognate RNA-DNA hybrids.

Conclusion

Through the combined application of special design strategies, computational tools, and developing production methods, recent advancements in RNA Nanotechnology have shown the creation of modular RNA units forming a wide variety of compact, thermodynamically stable structures for a broad range of applications. RNA architectonics has been used to rationally engineer tertiary assemblies into nanoparticles garnering increased resistance to nuclease cleavage as exemplified by molecular jigsaw-puzzles. Extensive studies of kissing-loop interactions and stabilization by tertiary folding multi-helix junctions have experimentally elucidated some of the thermodynamic constraints that dominate RNA folding.

While the potential use of functional RNA assemblies as packaging and delivery vehicles of therapeutic RNAs is presently in progress, the limitations in production of RNA nanostructures is immediately apparent through comparison to some naturally occurring RNA structures. At the moment, it is impossible to design RNA NPs with the size and functional complexity such as the ribosome, spliceosome and the pRNA motor of bacteriophage Phi29. These structures are far from discouraging, however; rather they provide a frame of reference with which to gauge the capacities of RNA as a platform. The ribosome can be used to demarcate a rough boundary line in terms of size and complexity. The bacteriophage Phi29 pRNA has proven that complex RNA structures can be redesigned for numerous uses, and suggests a future for more advanced RNA nanodevices and techniques. Similarly, the assembly strategy hypothesized for DsrA in bacteria, which is extremely similar to that of single stranded tile assembly, shows that the application of nucleic acid nanoscaffolds naturally extends in vivo.

While the vast majority of nucleic acid nanodevices are to date limited to in vitro and in vivo applications, this is changing. Synthetically designed RNAs have been expressed in living cells, and RNA nanoscaffolds which control spatial organization of biomolecules have demonstrated the ability to survive cellular conditions; recent work has shown that synthetic nanostructures can both assemble in vivo and function to optimize metabolic pathways.

Many challenges to realizing the potential of RNA nanotechnology remain, but with further developments in computational and design methods, combined with an ever growing library of knowledge of RNA structures and properties, we will be able to move toward the rational creation of increasingly complex, multidimensional architectures capable of performing myriad roles in nanomedicine, synthetic biology, and in the greater field of nanotechnology in general.