Elements of RNA Design

The versatility of naturally occurring RNAs that perform central cellular functions has generated intense interest in designing engineered RNAs to control these processes. (1,2) Mechanistic studies of natural RNA systems that sense and respond to changes in temperature, small molecule or metabolite concentration, and the expression of other RNA and protein species in the cell have been combined with advances in computational nucleic acid design to enable the de novo design of synthetic RNA regulators to control gene expression. While progress has been remarkable, (2−4) challenges remain. Here we outline important progress and emerging technologies for state-of-the-art RNA design and challenges which must be addressed to enable the next generation of synthetic RNA devices that can be used in applications including biomanufacturing, therapeutics, and diagnostics (Figure 1).

Figure 1

High Resolution Image

Download MS PowerPoint Slide

A primary consideration in RNA design is the RNA folding regime most appropriate for a desired function. Two principle RNA folding regimes are most important, depending on which point in its lifecycle the RNA performs its function. Kinetic design focuses on the out-of-equilibrium folding processes that occur during RNA synthesis, where interactions and refolding events must occur during the fast process of transcription. Examples of RNAs that need kinetic design include riboswitches that make transcriptional regulatory decisions as a function of intracellular ligand concentration. On the other hand, thermodynamic design focuses on RNA functions that can occur over longer time scales, and is typically appropriate when interactions occur post-transcriptionally and are driven by free energy differences between possible RNA structures and interactions.

Most work in RNA synthetic biology to date has focused on thermodynamic design, primarily because we know the most about equilibrium RNA folding. One example is the toehold switch, a computationally designed RNA system that controls translation in response to the binding of a designed RNA trigger. Toehold switches allow tight control of translation and have enabled the most complex RNA-based synthetic gene regulation to date that can implement the in vivo evaluation of a 12-input logic function. (4) Kinetic RNA folding pathways on the other hand are generally more difficult to computationally predict, and therefore, designing in the kinetic regime has remained a challenge for RNA synthetic biology. If this challenge can be overcome, however, the kinetic RNA folding regime has several potential advantages, including the ability to regulate transcription with designed RNAs that would enable RNA-only genetic circuits. For example, de novo-designed transcriptional activators called small transcription activating RNAs (STARs) have been used to create genetic networks that display sophisticated temporal patterns of gene expression, including an incoherent feed forward loop governed by the STAR-mediated control of guide RNA synthesis for downstream CRISPR interference. (3) Both of these examples are evidence that there has been strong progress in RNA synthetic biology, with computationally designed STARs and toehold switches achieving dynamic ranges exceeding those of many protein-based systems. (3,4)

Another element to consider is the design of RNA interactions with small molecules, ions, and proteins. These interactions allow natural RNA regulators to serve as biosensors of cellular state. However, ligand–RNA and protein–RNA interactions are difficult to model in silico, as design algorithms cannot currently capture the many noncanonical structural contexts often needed to create specific binding interactions. Recently, progress was made in the computational design of synthetic riboswitches by forcing the fold of the aptamer domain and accounting for the additional energy of stabilization from ligand binding, representing an important step forward in effectively modeling these complex RNA–ligand interactions in synthetic systems. (2) However, substantial work remains to enable the completely de novo design of such interactions, and will require advancements such as the integration of RNA and protein design algorithms, as well as the further development of strategies to abstract the complexity of RNA–ligand binding to generate simplified design motifs.

Dimensionality is also an important element of RNA design. Most in silico design methods have been limited to two-dimensional (2D) design and typically do not allow for the consideration of the three-dimensional (3D) tertiary interactions that can be critical for proper RNA function. Designing at the level of secondary structure is often a suitable abstraction to achieve successful designs, particularly when RNA function is based on the conditional formation of base pair interactions. (3,4) However, designing more complex regulation strategies will require improved folding models to account for the three-dimensional nature of RNA folding. Recently, there has been progress in bridging the gap between de novo 2D and 3D design, where structural motifs taken from solved crystal structures were used as building blocks to design complex 3D structures including linkers for synthetic tethered ribosomes. (5) This work provides a glimpse of the staggering breadth of novel functions that de novo 3D design could enable.

Following design, functional and structural characterization form the backbone of the testing platform for RNA engineering. Functional testing links expression of a designed RNA network to a measurable cellular output such as gene expression. However, it can be difficult to explain discrepancies between the designed structure of an RNA regulatory network and an undesired functional output. A unique tool for RNA-based design is the broad suite of high-throughput chemical probing technologies that have been developed to characterize RNA structures in a range of folding regimes including in vitro, in vivo, and cotranscriptionally folded. Chemical probes covalently modify an RNA of interest in a structure-dependent manner, with more flexible regions being more readily modified. The positions of these modifications are then detected by reverse transcription and high throughput sequencing. The distribution of modifications is then used to calculate a reactivity at each position, with higher reactivities corresponding to more unstructured positions. These experiments can capture structural features beyond simple base-pairing, such as protein or ligand binding, base stacking, and tertiary interactions. Since these features can all be difficult or impossible to predict with current RNA folding models, chemical probing experiments are an important bridge between in silico design and in vivo function, accelerating the design–test–learn cycle. In a recent example of this integrated approach, parallel functional and structural testing of the pT181 RNA transcriptional repressor revealed a new design principle related to the structural flexibility of an RNA–RNA interaction domain, enabling the forward engineering of new repressor variants. (1) A question at the forefront of RNA synthetic biology is how to best integrate high-throughput RNA structural information within design algorithms.

While much has been learned in recent years, several challenges remain for the creation of complex RNA networks. First, the burden imposed by synthetic circuitry can be deleterious to cellular fitness. Synthetic RNAs are typically overexpressed relative to endogenous RNAs in order to minimize diffusion limitations and maximize regulator performance. (3,4) For small networks, this overexpression does not have a substantial impact on cellular fitness, but larger networks containing multiple layers of regulation or many individual species can lead to drastically reduced cellular fitness. Strategies to obviate the need for overexpression, such as the development of synthetic organelle-like vesicles to increase the local concentration of interacting species, will be necessary to minimize cellular burden for complex genetic circuitry. Context effects, that is, the spatial ordering of individual functional elements along an RNA transcript, can also impede the functional performance of complex RNA regulatory networks. Elements that behave as expected in isolation may interact with adjacent functional elements when expressed as part of a larger network; these long-range interactions are difficult to predict a priori and currently require a trial-and-error approach to test which organizational schemes perform best. Isolation strategies, including the placement of cleavable RNA elements between functional domains, (3) can help reduce these effects, although improved structural design principles are still needed. Finally, limited portability between organisms remains a fundamental roadblock. Theoretically, RNA folding could be robust to different cellular environments, which is an intriguing potential advantage of RNA regulation. However, in practice, RNA regulators are generally optimized to function in a single host and rely on cellular machinery—such as RNA polymerases (3) and ribosomes (4)—for proper function. Transfer of regulatory mechanisms to nonmodel organisms has therefore been difficult, limiting the widespread adoption of these regulation strategies. Decoupling these strategies from organism-specific features of cellular factors could facilitate the efficient transfer of RNA regulators to new hosts.

Here we have outlined several elements of RNA design that have fueled significant progress in RNA synthetic biology. The second edition of these elements will likely include deeper insights that distill the nuances of the structure–function relationships of RNA into principles of RNA design. On the basis of the rapid and impressive progress in RNA synthetic biology thus far, we anticipate designed RNA systems to play an increasingly important role in biotechnologies and to become powerful synthetic tools for unraveling the function of natural RNAs.