Regulatory RNAs

Abstract

RNAs have many important functional properties, including that they are independently controllable and highly tunable. As a result of these advantageous properties, their use in a myriad of sophisticated devices has been widely explored. Yet, the exploitation of RNAs for synthetic applications is highly dependent on the ability to characterize the many new molecules that continue to be discovered by large-scale sequencing and high-throughput screening techniques. In this review, we present an exhaustive survey of the most recent synthetic bacterial riboswitches and small RNAs while emphasizing their virtues in gene expression management. We also explore the use of these RNA components as building blocks in the RNA synthetic biology toolbox and discuss examples of synthetic RNA components used to rewire bacterial regulatory circuitry. We anticipate that this field will expand its catalog of smart devices by mimicking and manipulating natural RNA mechanisms and functions.

Glossary

Composability: refers to the property of a system to break down in units (parts) due to the system modularity and ability to recombine in different configurations to satisfy specific human requirements.¹

Biological “part”: a genetically encoded functional unit that can be composed into a more complex assembly (e.g., a transcription promoter, an RNA aptamer, or a gene reporter).²

Genetic device: is an assembly of biological “parts” designed to execute a human-defined biological function.²

Regulator: biological part embedded in a regulatory-network and exerts constitutive control over gene expression. It determines the gene expression dynamic range.²

Sensor: biological part that detects a signal and transforms it into a gene-regulatory action or a structural reconfiguration of another biological part.²

Adaptor: non-coding nucleotide sequences or unstructured peptides connecting parts that isolate function by providing physical separation.²

Actuators: reporter genes that provide an output signal either as a particular phenotype or measurable outcome.²

Stabilizers: we define stabilizers as biological parts (e.g., non-coding sequences of nucleotides) that prevent decay of another part or device by providing physical protection and/or being docking domains for other helpers (e.g., Hfq domain in trans-encoded antisense sRNAs).

Introduction

The RNA world has been shaken by the recent recognition that small transcripts, once thought as “junk,”³ can serve critical roles as cellular regulators. This period of intense discovery has led to the realization of the ubiquitous presence of these short non-coding RNAs throughout bacterial genomes, and of their value as controllers.^4-6 Furthermore, regulatory RNAs are involved in important sensing functions and regulatory responses to environmental cues.⁷

As these sophisticated regulators are explored and their mechanisms and functions unveiled, Synthetic Biology has continued to arise as a field. Scientists now describe it as a discipline that pursues the engineering of life. As such, the ultimate goal of this field is to synthesize biological entities that can perform non-natural functions, in rational, standard, and systematic ways. Importantly, the discipline of Synthetic Biology currently spans from the biomolecular level to the macroscopic level of whole cells, tissues, and even organs.⁸ Some early examples include the genetic toggle switch⁹ and the repressilator.¹⁰ In this earlier stage promoters, ribosome binding sites, and repressors were at the core of the artificial manipulation work. Later on, these genetic control elements were combined to generate a diversity of devices with specific responses (for a review, see refs.¹¹ and ¹²).

Synthetic Biology has been empowered by a wide variety of technologies that include the ability to combine existing techniques that allow import and export of natural capabilities across different organisms. These technologies include traditional cloning, cloning of entire biological machines,¹³ fabrication of metabolic pathways,^14,15 alteration of the amino acid sequence of a protein,^11,16 design of tailored RNA molecules,¹⁷ codon-optimization,¹⁸ and chemical modification of biological entities to build upon natural regulatory mechanisms.¹⁹

The case for using RNA in Synthetic Biology is based on the ability of RNA elements to exert: independent control (particularly cis-acting molecules) by acting on a single molecule in a specific way, high tunability since their structures can be easily manipulated, orthogonality given the ability to prevent undesired interactions between variants, scalability due to their intervention in all levels of gene expression, composability given their relative potential to be assembled in a modular way, and portability since they can be more readily transferred to other organisms relative to their protein counterparts.²⁰ Another advantage of using RNA-based sensing approaches is that these molecules exhibit more compact genetic footprints than their protein counterparts and, as such, are more “economical” since they generally place less of an energetic load on the cell by not requiring translation.²¹ Furthermore, RNA-based post-transcriptional responses generally act at a faster time scale than transcription-based strategies²² and RNA-based mechanisms can be evolved more rapidly than protein-based mechanisms.²³

Since 1982, when the Nobel Prize committee recognized the discovery of RNA catalytic properties²⁴ as groundbreaking,²⁵ several “big” catalytic RNAs have been exploited for various biotechnological and medical applications. The catalytic RNAs that have made exciting headlines include group II introns,²⁶ RNase P,²⁷ the group I ribozyme,²⁸ and the Hammerhead catalytic module²⁹ (also known as the Hammerhead ribozyme-HHR). These applications represent some of the first developments that preceded small regulatory RNAs in the Synthetic Biology field. Our improved understanding of the design rules that govern these highly specialized regulatory molecules enables the visions of replicating their unique attributes and of further enhancing their roles. We also begin to see how these regulatory units could be transferred across living and non-living systems for the bottom-up synthesis of completely novel and highly controllable cellular phenotypes. Figure 1 is a general overview that illustrates the universe of RNA devices and parts, and the different levels of complexity that scientists have been able to reach by engineering RNAs in Synthetic Biology.

Figure 1. Universe of synthetic RNA devices and parts. An overview of the universe of RNA devices and parts that have been engineered in all three kingdoms of life. The red square delimits the space of RNA devices that this review focuses on. The blue arrow to the left of the picture represents how the complexity increases from bottom to top. The classification of RNA devices has been made consistent to previous work.¹¹⁹

In the remainder of this article, we survey recently reported and some pioneering mechanisms and applications that have been associated with riboswitches (a cis-acting RNA molecule that can control expression of its neighboring messenger RNA-mRNA) and trans-encoded small RNAs (sRNA) (a trans-acting molecule that regulates mRNAs by antisense mechanisms). We aim to present a comprehensive introductory review for the non-expert reader. We focus on these two types of RNAs because they are two of the most understood classes of regulatory RNAs. We also compare their natural mechanisms and molecular properties. And finally we offer insights into the types of opportunities that exist in this field to enable full exploitation of regulatory RNAs in Synthetic Biology as directed toward biotechnological applications.

Natural Riboswitches and sRNAs

Riboswitches: Single gene expression controllers

Riboswitches (Fig. 2) are natural genetic control elements that are found in cis at the 5′ (or 3′) untranslated regions of mRNAs. These regulatory RNA elements typically regulate expression of downstream genes by binding to cellular metabolites (e.g., small molecules) that cause a “switch” in their structural conformation. As a result, this structural switch leads to changes in expression levels of their partner gene.³⁰ Likewise, several riboswitch molecular architectures have been characterized;³¹ although these exhibit a certain degree of similarity in their local motifs, they show a diverse array of secondary and tertiary structures. For a recent review of riboswitches, see reference ³². The basic functional mechanisms for three of the main classes of riboswitches are illustrated in Figure 2.

Figure 2. Gene expression control mechanisms of ligand-binding riboswitches. (A) Transcriptional elongation control by a ligand-activated riboswitch. In the absence of ligand, an anti-terminator stem is present allowing completion of transcription by a polymerase. A ligand binds a specific site in the aptamer and causes the formation of a terminator stem that arrests transcription and prevents the synthesis of a complete transcript (dashed line). The opposite mechanism in which transcription turns on upon binding has been identified in nature to a lesser extent (not shown in figure). (B) Translational initiation control by a ligand-activated riboswitch. In the presence of the ligand, a riboswitch undergoes a conformational change and sequesters the Ribosome-Binding Site (RBS) to prevent the ribosome from binding the mRNA. In the same way as with transcriptional controllers, there are examples of riboswitches that turn on translation upon binding (not shown in figure). (C) Translational control by an allosteric ribozyme. The glmS glucosamine-6-phosphate (GlcN6P)-activated ribozyme (aptazyme) is shown. In the absence of GlcN6P, translation is ON and the transcript is protected against degradation. In the presence of the ligand binding to the aptamer, the transcript is cleaved in the 5′ UTR by RNaseJ1 and is subject to decay.³⁷

A variety of mechanisms have been uncovered for riboswitch regulation (reviewed in ref. 33). An interesting distinction is the one-domain (e.g., exhibited by thermosensors that respond to temperature changes) vs. two-domain molecular organization of riboswitches.³⁴ Most widely studied two-domain switches are RNAs composed of a ligand-binding aptamer and an “expression platform,” where the expression platform regulates gene expression in response to conformational changes in the aptamer region (Fig. 2A and B). Importantly, the aptamer region could be highly tuned for specificity in the types of molecules that it can recognize. One of the most powerful examples of this has been the ability to tune riboswitch recognition to discriminate between subtle structural differences in small molecules that differ only by the presence of an H and a CH₃ groups (e.g., theophylline vs. caffeine) with nM affinity.²¹ Indeed, it is the strong sensitivity intrinsic to these RNA switches that is highly exploitable in the detection of small molecules, macromolecules, and external stimuli. This extreme specificity is attributed to a modular component of riboswitches called the aptamer (Fig. 2). Aptamers are RNA molecules capable of binding a broad variety of small effector molecules and undergoing a structural reconfiguration. In fact, aptamers can be considered synthetic predecessors of natural riboswitches since they were evolved into conditional expression controllers in 1998,³⁵ several years before the first natural riboswitch was identified in the early 2000s.³⁶

A unique class of riboswitches is the so-called aptazymes. As shown in Figure 2C, aptazymes are allosterically controlled ribozymes. One example of this is the allosteric glucosamine-6-phospate (GlcN6P)-responsive ribozyme discovered by Winkler et al.³⁷ This aptazyme that is located at the 5′ UTR of B. subtilis glmS mRNA, cleaves and induces RNase J1 degradation of the message. The interest on this class of regulators is spotlighted by the fact that several aptazymes were engineered in vitro even before they were discovered in 2004.³⁸

An interesting one-domain type of riboswitch is the temperature and pH stimuli-responsive class, where RNA undergoes structural rearrangement in direct response to temperature or pH.^33,39 Although these RNA switches could be equally valuable for sensing technologies, they are far less studied and less exploited than metabolite-responsive elements. This is due in part to the paucity of riboswitches that have been uncovered in this category and to the lack of mechanistic insights and molecular principles of their function and design. Yet, depending on the application, these sensors can offer advantages over metabolite sensing elements since they can function without the need of an intermediate activator or an additional structural domain.⁴⁰ These features make these RNA switches more portable for certain synthetic constructions and are particularly attractive for applications in which using small effector molecules to trigger gene regulation is prohibitive.

One of the few reports geared toward engineering artificial thermosensors consisted of a structural model-based approach to design synthetic functional RNA thermometers⁴¹ capable of controlling translation initiation as efficiently as natural thermosensors. These temperature-sensitive riboswitches were engineered in E. coli using β-galactosidase as a gene reporter. In another example, Wieland et al.⁴² successfully engineered a series of temperature-sensitive riboswitches by a dual approach: GC-rich four-stranded-constructs (quadruplexes), suspected to be of relevance to the control of gene expression, were rationally designed and experimentally tested in vivo for suppression of a reporter gene. Quadruplexes of moderate secondary structure stability, as determined by circular dichroism spectroscopy, show temperature sensitivity in close resemblance to RNA thermosensors. Interestingly, in recent years, thermosensors have been treated in the literature almost as a separate class of regulatory RNAs. This is partly due to the far less intense exploitation of one-domain riboswitches for artificial gene management applications in contrast with other classes of riboswitches.⁴⁰ Therefore further discussion of riboswitches will focus on ligand-binding riboswitches (including aptazymes) and discuss the coupling of riboswitch-sensing ability with response circuits to generate new synthetic functionalities.

sRNAs: Multiple gene expression controllers

Small RNAs (sRNAs), ranging from 21~400 nucleotides, are an intriguing class of RNAs that (like riboswitches) do not encode proteins but have intrinsic roles as cellular regulators.⁴³ A key property of sRNAs is their ability to simultaneously turn on and off a variety of metabolic pathways in response to environmental changes (e.g., nutrient availability, osmolarity, pH, and temperature).⁴⁴ To exert their function, they can (1) base-pair with mRNAs (mRNAs) to enhance or attenuate transcription (Fig. 3A), (2) bind mRNAs to directly block (Fig. 3B, i), or indirectly enhance or inhibit translation (Fig. 3B, ii), (3) sequester proteins into ribonucleoprotein complexes to prevent their activity (not shown in Fig. 3), or (4) directly catalyze mRNA and protein degradation (Fig. 3B, ii).⁴⁵ The importance of sRNA regulatory mechanisms in bacterial survival and adaptation is evident by their ubiquitous presence in microbes.⁴⁶ Indeed, it is known that sRNAs can be essential for organism survival under stress.^43,44,46 Recent reviews on this topic highlight the diversity of sRNAs that continues to be evident with discovery of sRNAs in a wide range of bacterial species.^4,47

Figure 3. Gene expression control mechanisms by bacterial sRNAs.⁴⁵ (A) Transcription attenuation/enhancement. (A) sRNA binds to its target mRNA and causes a structural reconfiguration upon base-pairing, ultimately enhancing or attenuating transcription by the polymerase. (B) Translational control. Translational control is imparted by sRNAs in various ways: (1) A sRNA base-pairs to its target mRNA sequestering the Ribosome-Binding Site (RBS) and directly prevents translation initiation by the ribosomes. (2) A sRNA binds to the target mRNA at a distance from the RBS and the target mRNA suffers a structural change that indirectly affects ribosome binding. sRNA binding to its target can also enhance or inhibit mRNA decay by changing interactions with exonucleases and/or endonucleases.

Remarkably, sRNAs are also extremely precise in their recognition of RNAs and proteins; this is evidenced by the observation by the minimal basepairing requirement (8-9 basepairs) to bind a specific target.⁴³ The structural robustness and plasticity of small RNAs has recently been discussed in the context of these attributes’ evolutionary advantage. As it has been noted, these molecules can tolerate a large degree of “functional innovation” via acquired mutations without compromising their structures.⁴⁸

The fact that small RNA molecules can control major metabolic transformations has attracted much scientific interest. One of the most intriguing aspects of sRNAs is their ability to bind (and even sequester) multiple proteins (in addition to mRNAs), posing a rather powerful and unique regulatory mechanism. Two examples of regulatory protein-binding sRNAs that have been extensively studied are the CsrB/C (carbon storage sRNA regulatory system)⁴⁹ and 6S RNA.⁵⁰ Although regulatory RNA–protein interactions are only one form of global post-transcriptional gene expression control, the topology of their networks, and their mechanism of action are unique relative to other forms of post-transcriptional regulation (e.g., sigma transcription factors, repressor networks, and operon regulation by ligand-responsive riboswitches). The uniqueness of this form of regulation lies in the fact that these regulatory RNAs are not encoded close to their mRNA (or protein) targets, can recognize RNA at the sequence and structural level, can work by combining several binding domains in a modular way, and can directly control RNA stability by protection mechanisms.⁵¹

Bacterial small RNAs can be categorized as cis-encoded and trans-encoded. The former class refers to those encoded in a partial overlap with their target genes, whereas the latter are located in the genome apart from their targets. Since trans-encoded sRNAs account for the great majority of bacterial sRNAs, we will focus hereafter on trans-encoded sRNAs and will refer to them as sRNAs. Many trans-encoded sRNAs interact with multiple target mRNAs.⁵² In addition to these interactions with mRNAs, associations with the cellular chaperone Hfq have been widely observed, particularly in gram-negative bacteria. Hfq’s complexation with sRNAs is especially important in times of cellular and environmental stress.^47,52 Hfq is a major RNA chaperone that has been shown to: (1) increase the stability of small RNAs in vivo and in vitro; (2) bind several mRNAs and sRNAs to assist their pairing in global stress responses, (3) stabilize mRNA-small RNA interactions,⁴³ and (4) facilitate negative regulation by delivering the sRNA-mRNA pair to the degradosome (multi-ribonuclease complex).⁵² The cellular importance of Hfq is further underscored by the fact that it is a well-conserved protein in bacteria with homologs in eukaryotes. i.e. Sm proteins that form part of the spliceosome and play important roles in RNA metabolism.⁵³ It is, however, important to note that Hfq is not present in all bacterial organisms where sRNAs have been discovered.⁵⁴ In general, these findings suggest at least two possibilities: (1) that additional mechanisms (currently unknown) exist in these cells for sRNA stability purposes or (2) that not all bacterial sRNAs require protein chaperones for stabilization or assistance with target binding. Increased knowledge on natural protection mechanisms that have evolved in living systems could be highly useful in mimicking stability strategies for RNAs, particularly in the context of synthetic drug release vehicles. For recent reviews on Hfq function, structure, and its relationship with small regulatory RNAs, see references 7, 52, and 55.

Interactions with Hfq are intriguing from a Synthetic Biology perspective, given that their stability is highly relevant to making regulatory RNAs portable for applications outside the natural cellular environment. Although not much work has been done in this area, two general strategies for stabilizing RNAs have naturally evolved: special secondary structural motifs (e.g., hairpins and loops) and partnerships with proteins, such as Hfq.⁵⁶ For a recent review on this topic, see reference ⁵⁷.

In summary, important mechanistic differences exist between small RNAs and riboswitches (discussed above) that make them ideal candidates for different sets of applications. A gross description of these two “styles” of gene regulation can be captured by the thought of riboswitches being “local” managers, in stark contrast to the more “global” gene management style of sRNAs. Regarding applications, it is easy to imagine several cases where manipulation to gene expression could be beneficial at a highly localized level (without affecting other genes) for which synthetic riboswitches might be ideal candidates. However, sRNA’s ability to co-manage entire gene clusters and have molecules that could be engineered to be more promiscuous in their interactions also poses significant advantages in a context where broader molecular targeting is desired. Table 1 summarizes comparisons between natural features (with potential for Synthetic Biology applications) exhibited by riboswitches and sRNAs.

Table 1. Comparison of natural features exhibited by ligand-activated riboswitches and trans-encoded sRNAs.

Differences	Ligand-binding riboswitches	Small RNAs (trans-encoded)
Functional role	• Regulate single gene transcript	• Regulate multiple gene transcripts
	• When in tandem, exhibit complex gene control response	• Can be arranged in regulatory networks of high complexity
	• Triggered by a ligand	• Differentially expressed under certain stress conditions
Structural mechanism	• Do not need chaperones	• Trans-encoded could require Hfq
	• Undergo structural reconfigurations upon binding	• A diversity of mechanisms has been elucidated including target mRNA structural changes upon binding and indirect inhibition/enhancement of transcription; translation; and/or degradation
	• Direct triggering	• Indirect triggering: differentially expressed upon certain stress conditions (54)
	• Immediate response upon binding	• Potentially, concentration increase of Hfq upon environmental stress might increase sRNA activity
	• Single interaction between aptamer-metabolite activates response	• Multiple interactions: sRNA-Hfq, Hfq-mRNA and sRNA-mRNA base-pairing used to activate regulation
	• Highly specific pocket-like interaction with a small molecule	• High specificity for target mRNAs through reduced complementarity1
Structural	• One RNA motif: aptamer	• At least 2 RNA motifs: Hfq binding site and mRNA binding site.
	• Usually one hairpin-like structure	• Usually more than one hairpin-like motif
	• Attached to mRNA	• Independent, stable RNAs (could be co-transcribed with non-target mRNAs)

¹ Reduced complementarity: few nucleotides (8–9) used for base-pairing with target⁴³ in contrast with the eukaryotic counterpart. Also, often base-pairing is not perfect either, since it could include bulges. Note: sRNAs referred hereby are mRNA-binding molecules. Protein-binding class of sRNAs is not included here.

Recent Synthetic Regulatory RNAs for Engineering of Novel Cellular Capabilities

A considerable number of successes indicate the potential for using regulatory RNAs to expand our current Synthetic Biology toolbox and to help solve important problems in metabolic engineering, environmental remediation, agriculture, and molecular biology (Table 2, Table 3, Table 4, Table 5). Next, we review some of the most recent examples of these successes, with a focus on engineered sRNAs and riboswitches and their specific attributes.

Table 2. Recent synthetic riboswitches and aptazymes and their (potential) applications.

Application area (contribution)	RNA Unit Device	Function	Host	Description	Strategy	Reference
Environmental (bioremediation) Medicine (disease sites recognition)	Small-molecule responsive riboswitch	Translational initiation control	E. coli	Engineering cells to seek and destroy atrazine	In vitro evolved sensor (aptamer) In vivo evolved adaptor (expression platform)	58
Environmental (bioremediation) Medicine (disease sites recognition)	Small-molecule responsive riboswitch	Translational initiation control	E. coli	Engineering cells motility to navigate toward theophylline (predecessor of system above)	In vitro evolution of complete RNA device and tested in vivo	59
Synthetic Biology (RNA adaptor)	Slippage-like small-molecule responsive riboswitch	Translational initiation control	B. subtilis S. coelicolor	Engineering theophylline riboswitch with a new response mechanism	In vivo development of a RNA adaptor based on a slippage mechanism It was tested in S. coelicolor to confirm portability	121, 122
Synthetic Biology (RNA device)	Small-molecule responsive aptazyme	Translational initiation control	E. coli	Engineering Hammerhead ribozyme to be theophylline responsive and upon scission of mRNA turn translation ON	In vivo evolved RNA adaptor Natural scissile adaptor2 and sensor	111
Molecular Biology (small-molecule concentration detector)	Small-molecule responsive riboswitch	Translational initiation control	E. coli	Application of an adenosylcobalamin -sensitive natural riboswitch for the detection and study of B12 metabolism and transport	Natural sensor fused to different actuators	123
Metabolic Engineering (gene expression controllers)	Allosteric riboswitches (aptazymes)	Degradation-mediated gene expression control	E. coli	Model-driven engineered glmS-like aptazymes to control gene expression Additional contribution: computer-assisted design method	In vitro, in vivo and in silico engineered sensor and scissile adaptor2 assembled into catalytic regulator	60
Synthetic Biology (new RNA device)	Small-molecule responsive riboswitch	Translational control	E. coli	Reversed natural thiamine pyrophosphate riboswitch by dual genetic selection	In vivo evolution of adaptor	124
Synthetic Biology (a “universal” riboswitch)	Small-molecule responsive riboswitches	Gene expression control	E. coli A. baylyi A. baumannii A. tumefacians M. magneticum M. smegmatis B. subtilis S. pyogenes	Engineering 5 synthetic riboswitches that show gene expression control over 8 different bacterial species	Rational design and in vivo screening of ligand-sensitive regulators	125
Molecular Biology (Study of essential genes through hypomorphic mutants)	Small-molecule responsive riboswitch	Translational initiation control	E. coli	The natural theophylline riboswitch was used to control gene expression in hypomorphic mutants to study CsrA1	Rational design and in vivo screening of ligand-sensitive regulators	126
Medicine and molecular biology (riboswitch-based system to control gene expression)	Small-molecule responsive riboswitch	Translational control	M. Tuberculosis	Transcription promoter coupled to synthetic theophylline riboswitch for expression control	DNA transcriptional regulator coupled to a synthetic ligand-sensitive regulator	127 This is a continuation of ref.58
Molecular Biology (gene regulation and intracellular sensor)	Small-molecule responsive aptazyme	Translational control	E. coli	Changed aptamer from a previously engineered aptazyme to a theophylline-sensitive one	Interchangeable sensor coupled to a scissile adaptor2	128
Molecular Biology (intracellular sensor)	Small-molecule responsive aptazyme	Translational control	E. coli	Co-factor recognition aptamer fused to a Hammerhead ribozyme and anti-Ribosome Binding Site (RBS) sequence System detected enzyme co-factor concentrations	Interchangeable sensor coupled to a scissile adaptor2	129
Molecular Biology (biodetectors)	Small-molecule responsive riboswitch	Translational control	E. coli	Explored the applications of fusing synthetic riboswitches to gene reporters as a genetic screening method	In vitro evolved synthetic ligand-responsive regulator	130

¹ CsrA: Carbon Storage Regulator “A” protein. ²Scissile adaptor: an RNA, DNA or protein sequence that is cleaved for the part that it connects to, to perform its function (e.g., expression platform in aptazymes). ³Amber suppression: a tRNA has been modified to recognize the amber stop codon and instead insert an amino acid thus suppressing translation termination.¹³¹

Table 3. Recent synthetic sRNAs and their (potential) applications (basic devices).

Application area (contribution)	RNA Unit Device	Function	Host	Description	Strategy	Reference
Molecular Biology (RNA intracellular levels detector)	Protein-binding sRNA	Transcriptional regulation	E. coli	Engineering small non-coding RNA for the expression control of TetR (a bacterial repressor)	Stabilizers1 added to an in vitro evolved sensor at each end and the assembled regulator was screened in vivo	132
Metabolic Engineering (strain optimization)	Synthetic Small trans-acting RNAs	Gene expression control	E. coli	Engineered Synthetic sRNAs for gene combinatorial knockdown and target identification	In vivo selected regulators were engineered by using MicC2 Hfq- binding domain as a scaffold A target-binding domain was rationally designed	68
Metabolic Engineering (gene knockdown)	Synthetic Small trans-acting RNAs	Gene expression control	E. coli	RecA3 knockdown using synthetic sRNAs	In vivo screen of regulators with randomized target-binding sequence	133
Synthetic Biology (computational approach to the design of artificial sRNAs)	Synthetic Small trans-acting RNAs	Gene expression control	E. coli	Artificial sRNAs computationally designed using an energy-based optimization model that could be expanded to the rational design of other devices	Engineering regulators by a computational approach that includes energy-based models Later regulators were experimentally validated	134
Molecular Biology (tool to study sRNA gene-silencing mechanisms)	Synthetic Small trans-acting RNAs	Gene expression control	E. coli	Investigation of sRNA gene-silencing mechanisms by using artificial sRNAs	Rationally designed regulators by modifying target-binding domain and in vivo testing using an actuator	135
Synthetic Biology (knowledge for design of artificial sRNAs)	Synthetic sRNA-like antisense devices	Translational initiation control	E. coli	Shine-Dalgarno sequence plays a main role in the antisense gene-silencing mechanism of artificial antisense short RNAs	Rationally designed and experimentally tested regulator by modifying anti-Shine-Dalgarno sequence region	136
Synthetic Biology (method for designing sRNAs)	Synthetic Small trans-acting RNAs	Gene silencing	E. coli	sRNAs engineered by rationally introducing two binding domains and a transcription terminator	Rationally designed regulator tested in vivo	137
Synthetic Biology (new orthogonal devices and computational tool)	Artificial antisense sRNA-like orthogonal pairs	Translational regulators	E. coli	Computational model-driven design of antisense RNA-RNA orthogonal pairs and experimental validation	Rationally designed antisense pairs of RNA regulators	138
Synthetic Biology (computational method to design artificial sRNAs)	Synthetic Small trans-acting RNAs	Gene silencing	E. coli	Energy-based computations directly correlated with a continuum of repression levels when modifying RyhB4 sequence	Multiple RyhB4regulator variants were found by introducing sequence variations in the target-binding site	104
Synthetic Biology (strategy to design conditional gene silencers)	Synthetic Small trans-acting RNAs	Gene silencing	E. coli	A high-throughput screening method was developed to engineer artificial sRNAs Two active synthetic sRNAs were found	Antisense regulator engineered from in vivo screening by randomizing binding domain and preserving scaffold	72

¹Stabilizers: we define stabilizers as biological parts (e.g., non-coding sequences of nucleotides) that prevent decay of another part or device by providing physical protection and/or being docking domains for other helpers (e.g., Hfq domain in trans-encoded antisense sRNAs) ²MicC: bacterial sRNA that posttranscriptionally regulates expression of ompC, a gene that encodes for an outer membrane protein.⁴³RecA: a relevant DNA-repair protein. ⁴RyhB: is a sRNA which acts to reduce iron consumption under low-iron conditions by downregulating expression of iron-containing proteins, including enzymes of the citric acid cycle and the aerobic respiratory chain.⁴

Table 4. Dual and chimeric riboregulators.

Application area (contribution)	RNA Unit Device	Function	Host	Description	Strategy	Reference
Synthetic Biology (RNA device)	A riboswitch-activated sRNA (chimera)	Transcriptional elongation and translational initiation control	E. coli	A natural theophylline aptamer activates transcriptional and translational regulators (sRNAs)	Natural sensor+adaptor coupled to a natural sRNA by engineering a kissing-loop2 –like interaction.	75
Synthetic Biology (Increased signal propagation speed for complex circuits)	A dual sRNA-like system	Transcription control	E. coli	Showed “composability” of these devices for a complex circuit by using variants of wt pT181 regulation system	Regulator and sensor interactions engineered by introducing rational mutations	74
Molecular Biology (robust gene expression regulation tool) Synthetic Biology(programmable kill switch)	A dual sRNA-like system	Translational initiation control	E. coli	Engineered riboregulator system (cr-taRNA) with a broad dynamic range	Rationally designed a diverse pool of regulators and tested in vivo	73 Potential applications ref. 139
Synthetic Biology(addition to platform/tool box)	An anti-sense RNA responsive aptazyme	Translation initiation control	E. coli	Trans-acting responsive element engineered into a HHR for construction of more complex genetic circuits	Rationally designed regulator-sensing device coupled to a scissile adaptor and an actuator.	76

¹ tnaC: is a regulatory element that contains a Ribosomal Binding Site, short leader peptide code and a rho terminator.¹It is based on the regulatory element from tna operon in E. coli.²Kissing-loop interaction: is a common coaxial stacking motif in which nucleotides in two loops of two different hairpins base-pair to form a stable tertiary structure.

Table 5. Recent examples of complex genetic circuits constructed by composing RNA regulators.

Application area (contribution)	RNA Unit Device	Function	Host	Description	Strategy	Reference
Synthetic Biology (band-pass RNA device)	Small-molecule responsive riboswitches in tandem	Transcriptional and translational control	E. coli	Coupled TTP-OFF riboswitch with another TPP-ON riboswitch in tandem	Two independently in vivo evolved regulators in tandem	62
Synthetic Biology (RNA device that performs Boolean logic operations)	Small-molecule responsive riboswitches in tandem	Transcriptional and translational control	E. coli	Logic gates AND and NAND1 were engineered by combinations in tandem of a theophylline aptamer fused to thiamine pyrophosphate	Two regulators in tandem independently obtained from a dual genetic screen	61
Synthetic Biology (RNA adaptor and two devices)	Antisense-inactivating sRNA-like and dual riboregulator (sRNA-like system)	Transcriptional elongation control	E. coli	Use and validation of tnaC2 as a universal converter from translational to transcriptional regulators	Artificial translational initiation regulators coupled to a new adaptor to engineer a transcriptional control device	1
Synthetic Biology (biocomputers3)	Dual riboregulator (sRNA-like system)	Translational control	E. coli	Riboregulator engineered previously75 used in two genes tied in a transcriptional cascade	Regulators arranged into a transcriptional cascade	78
Synthetic Biology (addition to platform/tool box)	Dual riboregulator (sRNA-like system)	Translational initiation control	E. coli	Using same device developed previously75 two complex genetic circuits were constructed: one for multi-sensing and another for metabolic regulation	Rationally designed and in vivo tested diverse RNA regulators variants and arranged in a “switchboard”	79
Synthetic Biology (Boolean logic operations)	Small-molecule responsive aptazyme	Translational control	E. coli	Translational control by synthetic aptazymes coupled to amber suppression3 generates logic gates	Ligand-binding regulators arranged into complex genetic circuits	140
Metabolic Engineering (expression of recombinant proteins)	Small-molecule responsive riboswitch	Transcriptional elongation control	B. subtilis	Using the natural glycine tandem riboswitch as a potential gene expression controller	Natural riboswitches fused to actuators tested in vivo	83

¹ A logic gate is an imaginary or physical device that performs a Boolean operation. NAND is a logic gate in which the outcome is false only if all the inputs are true. In contrast AND performs a Boolean operation in which both inputs should be true in order for the output to be true as well. ²tnaC: is a regulatory element that contains a Ribosomal Binding Site, short leader peptide code and a rho terminator. ¹It is based on the regulatory element from tna operon in E. coli.³Biocomputer: is a DNA, RNA and/or protein-based system capable of performing computational calculations including storing, processing and retrieving data.

Riboswitches: Important parts in the plug-and-play approach to synthetic biology

Most recently, remarkable scientific achievements have been accomplished in cellular and molecular engineering by our ability to program “intelligent” responses using synthetically designed artificial transcription–translation circuits.¹² In this context, riboswitches have played an important role as inspiring models for the engineering of living systems. They have been exploited as predictable, robust, and precise devices that can act as sensors of various environmental signals,^58,59 as metabolic engineering tools for gene expression control,⁶⁰ and even as biocomputers (logic operations and band pass filters).^61,62 General strategies to synthesize artificial riboswitches are shown in Figure 4, where we illustrate that these regulatory elements are made of decomposable elements.

Figure 4. Composability of riboswitches as a strategy for the synthesis of artificial RNA devices. Composability refers to the ability of a system to break down in units (parts) due to the system modularity and recombine in different configurations to satisfy specific human requirements. (A) Composability of riboswitches. Riboswitches can be decomposed and recombined for the synthesis of new devices with high modularity. An artificial riboswitch-based device is composed of a regulator (riboswitch), a signal (ligand), and an actuator (gene reporter). The regulator (inside the solid square) can be further decomposed into a sensor (aptamer) and an adaptor (expression platform that usually contains a terminator stem). (B) Composability of aptazymes (allosteric ribozymes). A synthetic catalytic device is composed of a catalytic regulator (aptazyme), a signal (ligand), and an actuator (gene reporter). The catalytic regulator (inside the solid box) can be further disassembled in a sensor (aptamer) and a scissile adaptor (ribozyme).

As aforementioned, aptamers have contributed to pioneer the field of engineered riboswitches since the idea of an RNA switch had already been devised even before natural riboswitches were discovered in the early 2000s.⁶³ Although aptamers have been obtained through combinations of in vivo and in vitro selections, they are typically developed by in vitro selection schemes such as Systematic Evolution of Ligands by EXponential enrichment⁶⁴ (SELEX). To date, this remains one of the most widely used and effective methods. Many different aptamers have now been developed and used as basic riboswitches in higher organisms (e.g., yeast and mammalian cells) or have been linked to more sophisticated engineered expression platforms (modular component that interprets and exerts the gene expression control action) (Fig. 2).⁶⁵ In addition, aptamers have been used as tools in a myriad of biotechnological applications both in vitro and in vivo; this further asserts their role as predecessors of riboswitches. Biotechnological aptamer applications include functional molecular biology studies, gene expression control, antibiotic research, diagnostics, biosensors, therapeutic agents, and drug targeting. For reviews of these systems, see references 66 and 67.

As for some riboswitch-based applications, specific eye-opening examples include the ligand-responsive riboswitch aimed for the search and destruction of the widely used agricultural herbicide atrazine.⁵⁸ In this example, artificial chemotaxis have been engineered in E. coli cells to navigate toward high concentrations of atrazine by the action of an in vitro- and in vivo-evolved atrazine-responsive riboswitch (Table 2). Specifically, an atrazine catabolic pathway was introduced for these artificial bacteria to destroy the herbicide. The RNA device was designed by taking advantage of the high modularity and composability of riboswitches (Fig. 4). This is one of the most developed applications and it reveals the potential of a synthetic riboswitch in the field of bioremediation and agriculture (Table 2). This same approach shows potential for rewiring of bacteria to find disease sites.⁵⁹ Table 2 lists a comprehensive survey of synthetic riboswitches in the field.

Exploiting sRNA regulation in metabolic engineering

Although many examples of designer sRNAs are being developed for a single target (Table 3), the natural ability of small RNAs to target multiple mRNAs can be highly exploited to tune a higher number of functionally interdependent genes and pathways for strain optimization.⁶⁸ In addition, sRNAs show high modularity due to their molecular makeup. These properties of sRNAs have empowered synthetic biologists to decompose and recombine sRNA parts to engineer artificial riboregulators with different functions. A diversity of strategies have been used to design and engineer the different sRNA parts for various applications that include rational design, model-driven computational design, in vivo and in vitro molecular evolution and selection and, even harvesting of natural parts. In general, the trend observed in a number of studies is the preservation of Hfq-binding domain and of a natural sRNA transcription terminator (both domains constitute the stabilizer) as a scaffold combined with the synthesis of remaining parts to be assembled on top of this template. Figure 5 shows a schematic representation of sRNA composability as a strategy for the synthesis of artificial sRNA devices.

Figure 5. Composability of sRNAs as a strategy for the synthesis of artificial RNA devices. Composability, as for the case of riboswitches (Fig. 4) is the ability of a system to break down in units (parts) due to the system modularity and recombine in different configurations to satisfy specific human requirements. sRNAs are regulators of high modularity. An sRNA-based regulator can be broken down in two main parts: a sensor (target binding domain) and a stabilizer (that can include an Hfq-binding site and the transcriptional termination domain). In the context of a genetic device, the sRNA binds an mRNA target. In this case, the 5′ UTR of the target mRNA acts as an adaptor that transmits the signal to the gene reporter actuator. The combination of the sRNA and mRNA target comprises a functional synthetic device.

The potential use of the capability to manipulate entire RNA-driven stress response circuits for cellular engineering has been previously exemplified by introducing gene knockouts within the carbon storage regulator (Csr) system. The Csr is a sRNA-driven system involved in bacterial responses to carbon depletion.^49,69,70 Specifically, a major sRNA component of this system (CsrB) has been targeted to increase intracellular levels of the phosphoenolpyruvate (PEP) metabolite, by diverting carbon flux from glycolysis to gluconeogenesis.⁷¹ Most recently, the implementation of synthetic small regulatory RNAs in metabolic engineering has been reported in the development of a sRNA-based system repressing translation of DsRed2 (red fluorescent protein) mRNA at various levels and confirming tunability by the modulation of lacZ (mRNA encoding for β-galactosidase) expression.⁶⁸ In this study, the construction of three different sRNAs was shown for the mRNAs LuxR, AraC, and KanR without cross-activity. This strategy enabled the isolation of E. coli strains optimized for the production of tyrosine and cadaverine. These results suggest that specific gene expression can be fine-tuned by the design of artificial sRNAs as an alternative strategy to gene knockouts. An advantage of this strategy is that synthetic RNAs alleviate the need to generate strain libraries and provide a wider range of gene expression.

Numerous factors likely influence sRNA efficacy when used in artificial regulatory schemes. Some of these include the types of mRNA targets that it can bind, kinetics of binding, and the extent of the sequence area that should be involved in binding. As a result, recent work has shown the notion of complementing rational sRNA design with high-throughput screening strategies that can efficiently identify synthetic RNAs that are capable of regulating genes in trans. A specific recent example is the strategy proposed by Yokobayashi and collaborators in 2011 for the design of artificial sRNAs.⁷² A high-throughput screening method was developed for the identification of synthetic sRNAs that efficiently represses endogenous genes in E. coli by arbitrarily selecting the well-characterized sRNAs (DsrA, GcvB, MicF, and Spot42) to test their methodology. In this work, the antisense domain was randomized and fused to sRNA scaffolds, preserving the Hfq binding domain unmodified. A library of random artificial sRNAs was generated and tested against a native 5′ mRNA leader sequence fused to Green Fluorescent Protein (GFP) as a reporter, leading to the finding of sRNA synthetic candidates that successfully repress two mRNAs (ompF and fliC). Interestingly, they observed key characteristics in those artificial constructs repressing ompF of high similarity to the features shown by MicF (known to regulate the same mRNA).⁷² In general, the strategy of maintaining a stabilizer (e.g., sRNA scaffold containing the Hfq-binding domain and a transcription terminator sequence) and reengineering the sensor (i.e., a target-binding domain) through a combination of rational and computational model-driven approaches seems to be commonly observed across recent works in this area. The same pattern is observed with the common application of in vitro and in vivo screening technologies in the field. These features are illustrated in Figure 5. We present in Table 3 an extensive survey of the synthetic sRNAs and of some of their potential applications.

As suggested by Tables 2 and 3, the types of strategies used to build synthetic devices and their applications are significantly different when using riboswitch and sRNA-like features. Riboswitches have been used in general for tighter local gene expression control, typically relying less on rational design and more on in vitro and in vivo evolution techniques, and mostly aiming for bio-sensing applications. On the other hand, sRNAs have been used mostly to build global artificial regulators for metabolic engineering applications. Engineered sRNAs seem to be more recent than their riboswitch counterparts and have in general relied more on rational criteria for their design and construction. Moreover, several sRNA-related studies have addressed the necessity of using modeling and in silico approaches to aid the synthesis of artificial sRNA-like regulators.

Engineering dual and chimeric riboregulators

In another class of applications, some groups have used a combined approach by engineering the cis-acting region of the target mRNA as well as the trans-acting antisense RNA (sRNA). Through this more holistic approach, de novo RNA devices have been created. For instance, Isaacs et al. have used this approach to construct a riboregulator that controls GFP expression with a high dynamic range. A cis-repressive element (crRNA) was designed to act as an anti-RBS that blocks ribosome binding. A trans-acting element (taRNA) transcribed from another promoter interacts with the crRNA-RBS stem-loop disrupting the structure and exposing the RBS to generate fluorescence by GFP expression (Fig. 6A).⁷³ This approach showcases the high composability of regulatory RNAs, further exemplified in Table 4. For instance, the use of an antisense small trans-acting RNA to induce transcription termination via the exposure of an intrinsic terminator stem in the transcription attenuator (from the plasmid pT181) has been illustrated in Figure 6B.⁷⁴ A similar creative way to combine the cis-acting capabilities of an aptamer to activate a small RNA is shown in Figure 6C.⁷⁵ This chimeric device boldly combines riboswitch-like features with sRNA-like features emphasizing the high composability of RNA devices. Finally, one of the most holistic examples is the engineering of an allosteric ribozyme susceptible to activation by a small antisense RNA that ultimately turns translation off (Fig. 6D).⁷⁶

Figure 6. Dual and chimeric riboregulators. Based on the composability of these regulatory molecules, they can be recombined and assembled into a new regulator that exerts a completely different function. (A) Translational control by a small trans-acting RNA-responsive switch. Isaacs and collaborators designed a riboregulator combining a riboswitch-like sensor (YUNR motif), an adaptor that contains an anti-RBS sequence (crRNA) that can pair with the RBS and an sRNA-like effector molecule (taRNA). Translation is OFF (left) when the taRNA absent since crRNA sequesters the Ribosomal Binding Site (RBS). Translation turns ON (right) in the presence of the taRNA since it releases the RBS by recognizing the YUNR motif¹²⁰ that leads to disruption of the CrRNA-RBS stem-loop.⁷³ (B) Transcriptional control by a small trans-acting RNA-responsive switch. A transcriptional attenuator (from plasmid pT181) was engineered to be RNA sensitive. The sensor loop in the attenuator was evolved to detect a short antisense RNA (that acts as an effector molecule) via a kissing-loop interaction. This interaction exposes an intrinsic transcriptional terminator (the adaptor) to turn transcription OFF.⁷⁴ (C) A ligand-responsive riboswitch activating an ncRNA. A chimeric regulator was engineered by using a natural aptamer as sensor and a non-coding antisense RNA (ncRNA) as the adaptor. By means of a kissing-loop interaction in the absence of the ligand, the aptamer interacts with the ncRNA hairpin inactivating its antisense function. In the presence of the ligand, the hairpin interaction is disrupted and the ncRNA recovers its regulatory functions.⁷⁵ (D) An sRNA-responsive aptazyme. A trans-acting RNA-sensitive hallosteric ribozyme was engineered by designing a sensor (TR: trans-acting RNA responsive element) that detects the trans-acting RNA and coupling it with a Hammerhead ribozyme (HHR). In the absence of the trans-acting RNA (left), the aptazyme undergoes cleavage exposing the (initially occluded) ribosomal binding site and allowing translation. In contrast, upon binding of the trans-acting RNA to the TR element, the Hammerhead ribozyme undergoes a structural change that renders it catalytically inactive; this masks the RBS and prevents translation initiation.⁷⁶

The construction of more complex, chimeric, and dual devices can be understood as the combination of properties exhibited naturally by riboswitches and sRNAs. This represents strong evidence of the RNA modularity. As illustrated in Figures 4 and 5, RNA parts include riboswitches and sRNAs (regulators) that can be further decomposed into aptamers and sRNA target-binding domains (sensors), riboswitch expression platforms and the 5′ UTR’s of sRNA target mRNAs (adaptors), Hfq-binding, and transcription terminator domains (stabilizers), among other parts.

Composing RNA regulators into complex genetic circuits

It is fascinating and almost surreal that decomposing and recombining RNA parts to construct advanced devices with novel functions is now a reality; some of the recent examples that represent this level of advancement in synthetic constructions are listed in Table 5. A particularly interesting example is the design of complex genetic circuits to perform Boolean logic operations based on the premise that the cell is by itself a biocomputer with the ability to perform complex closed-loop control over its own operations.⁷⁷ One research group for example, has arranged RNA-sensitive switches (crRNA-taRNA⁷³ Fig. 6A) in a complex network for elegant translational control with a high and robust dynamic range.⁷⁸ In addition, the design of Friedland and collaborators (illustrated in Fig. 7B) has the ability to count up to three events based on a system of three genes regulated by the crRNA-taRNA translational control system arranged in series. These three genes (T7 RNA polymerase, T3 RNA polymerase, and the gene reporter GFP) are controlled by promoters in a transcriptional cascade array. The taRNA is under the P_BAD promoter control, which is induced by arabinose. In the absence of arabinose, the system generates baseline levels of fluorescence. However, when pulses of arabinose are fed into the system, the fluorescence response highly correlates with the number of these arabinose injections. In another recent example, the same crRNA-taRNA riboregulation system⁷³ (Fig. 6A) was composed into a genetic circuit of high complexity, resembling a switchboard.⁷⁹ This circuit consists of an orthogonal parallel array of riboregulators that can be applied to multi-sensing tasks. When arranging the orthogonal pairs of crRNA and taRNA under the same promoter, the system gains the ability to monitor and regulate several metabolic pathways simultaneously (Fig. 7C). Liu et al.¹ developed an adaptor (Fig. 7D) that converts a translational regulator into a transcriptional regulator. To test their system, the researchers used the crRNA-taRNA riboregulator⁷³ and the IS10 antisense regulator.⁸⁰ The adaptor is a regulatory element based on the tnaC, a leader peptide from the tna operon in E. coli (for more details about the mechanism by which tnaC acts, refer to ref. 81). Finally, in a series of attempts to mimic natural systems, small-molecule riboswitches have been arrayed in tandem to constitute logic gates (Table 5) (e.g., NOR, NAND, AND, etc.).^61,62,82,83 This is another way by which the power of turning simple riboswitch-like RNA devices into complex translational and transcriptional control systems has been showcased.

Figure 7. Application of a translational regulator to construct complex genetic circuits. The following three examples use the crRNA-taRNA, a riboregulator that controls translation initiation (Fig. 5A, Isaacs et al.,⁷³ 2004), to engineer different synthetic systems. (A) An adaptor that converts regulators from translational to transcriptional. By fusing an adaptor to a translational regulator, the translational regulator becomes transcriptional regulator. The adaptor consists of a Ribosomal Binding Site (RBS), an Open Reading Frame (ORF) of a short leader peptide (tnaC), and a transcription termination region. By fusing this element to the translational regulator (e.g., crRNA-taRNA), translation of the adaptor is controllable (not shown). In the absence of the adaptor, regulation of GFP is at the translation initiation level by RBS sequestration. In contrast, by merging the adaptor, regulation of GFP is at the transcriptional elongation level; in this case, when the upstream translational regulator turns off synthesis of the adaptor, the transcription terminator in this element is exposed and arrests transcription of GFP (at the mRNA level). Likewise, when the translational regulator turns on synthesis of the adaptor, translation of GFP is activated at the protein level.¹ (B) Riboregulated Transcription Cascade (RTC) with the ability to count. By arranging 3 genes controlled by the translational regulator (crRNA-taRNA) into a transcriptional cascade a genetic circuit is built with the ability to count. When P_BAD promoter is not induced, the crRNA element present in each gene blocks translation of the T7 and T3 RNA Polymerase (RNAP), and GFP. In contrast, when P_BAD is induced with arabinose, the level of fluorescence generated correlates with the number of pulses of arabinose fed into the system.⁷⁸ (C) Switchboard for multi-sensing and metabolic pathway control. A series of rationally designed orthogonal variants of the crRNA-taRNA system (cr1, cr2…crn; taRNA1, taRNA2…taRNAn) are fused to multiple gene reporters (Reporter 1, Reporter 2…Reporter n) for simultaneous and independent regulation. The complex genetic circuit is arranged in parallel (each pair of crRNA-taRNA variants is under the control of a different promoter) to be able to sense multiple inputs and convert them into a measurable output (e.g., enzymatic and/or fluorescence reads).⁷⁹

Current Challenges and Opportunities for Synthetic Biology of Bacterial Regulatory RNAs

There is a need for additional strategies to assay natural capabilities of newly discovered RNA regulators to allow full sorting of the “hidden” wealth of information within all newly identified sequences. While recent high-throughput technologies^84-86 have enabled the fast discovery of a vast number of potential RNA regulators in bacteria, our ability to characterize their targets and structures lags far behind. This illustrates the gap between regulatory RNA discovery and mechanistic analysis. Below, we summarize current techniques that are typically involved in the functional characterization of regulatory RNAs. Although several have been successful, there is a need to develop efficient high-throughput strategies to catalog newly identified regulatory RNAs given the large volumes of data that are being produced in this area on a daily basis. In fact, up to recently, there was still a lack of consistency in the nomenclature used for newly discovered regulatory RNAs, making it difficult to keep track of the ones that were already identified (and perhaps even characterized) in the literature across research groups.

Experimental and bioinformatic strategies for target identification

Identification and complete validation of the cellular targets affected by regulatory RNAs is key to trace their mechanistic roles. Currently, experimental methods for sRNA target identification are numerous and of great variety as they are tailored to a regulatory RNA in question. Consequently, Vogel et al. have classified these strategies as genetic-based, chromosomal reporter gene fusions, affinity based assays, co-immunoprecipitation, proteomics, and translational fusion on plasmid.⁸⁷ Given the proximity of riboswitches to their targets, identification of their gene target is less challenging than in the case of sRNAs where the target can be encoded from anywhere in the genome. It is important to note that relying solely on individual strategies often favor identification of either protein or mRNA target, but not both. The challenge of functionally characterizing regulatory RNAs has been extensively reviewed in references 87 and 88.

Given the lack of uniformity in experimental techniques and the labor intensity of these methods, computational-assisted algorithms have become instrumental in this field. The use of bioinformatic approaches has been particularly helpful to define an initial pool of target candidates that can then be experimentally validated.^89,90 A successful example of the use of bioinformatic approaches for target prediction has been the identification of the sRNA OxyS target, the ybaY mRNA.⁹¹ Computational algorithms for these purposes have also been reviewed in references 91 and 92. In addition, recent innovative ways of characterizing sRNA have relied on a biology network approach to infer regulatory relationships between sRNAs and mRNAs.⁹³

A creative approach to bridge the gap between discovery of bacterial regulatory RNAs and functional characterization is to use a holistic whole cell approach. To this end, Modi et al.⁹³ proposed a network biology approach in which a computational algorithm was applied to the study of sRNAs. They fed extensive gene expression profiles into the algorithm to infer regulatory relationships between IsrA, GlmZ, GcvB, and their mRNA targets that were experimentally confirmed. The study provided strong support to the applicability of a regulator network approach to perform functional analysis on bacterial small RNAs while surpassing the accuracy and ease of current target prediction methods. Further advancement in our ability to place the function of a regulatory RNA within the context of the rest of the cellular environment is critical for the design of de novo RNA structures. This is particularly relevant to Synthetic Biology given that target binding specificity is another important metric in the functional evaluation of synthetic regulatory molecules that can fully exhibit the functional range of their natural counterparts.

Experimental and computational strategies for structural characterization

Successful design of desired synthetic functions into natural RNAs or de novo scaffolds is highly dependent on the ability to engineer targeted structural changes into these molecules. As such, structural characterization is an important step in the design of synthetic RNA performance. To date, X-ray crystallography and NMR are some of the most widely used techniques for determination of RNA structure. There are several examples of riboswitch structure and riboswitch–ligand complexes that have been studied using X-ray crystallography techniques.^94,95 However, these methods are limited mainly in terms of the molecular size ranges that they can handle. As a result, small-Angle X-ray Scattering (SAXS) and some adaptations of NMR techniques have been applied to resolve non-coding short RNA structures. Thus far, their applications have been mainly aimed at capturing dynamic structural reconfigurations in riboswitches and small RNA–Hfq interactions.^95-97

A diversity of chemical, biochemical, and enzymatic structural probing techniques has also been applied.^98-100 These techniques have been applied, either alone or in combination, to resolve structural features, conformations, and interactions in sRNAs and riboswitches. For instance, the binding of the RybB sRNA to the outer membrane protein OmpN mRNA has been resolved at the nucleotide level using enzymatic and chemical structural probing approaches.¹⁰¹ RybB is the sRNA that controls expression of outer membrane proteins (OMPs) in enterobacteria. There is a large body of published work reporting success with these techniques to characterize changes in sRNAs and riboswitches upon target binding. Recently, interactions of the glmS aptazyme with its effector molecules in B. subtilis were further investigated using enzymatic protection assays, leading to striking evidence that suggested a differential modulation of this riboswitch by multiple molecular cues.¹⁰² Another fascinating example is the recent development that couples enzymatic assays with transcriptome-wide sequencing to reveal the secondary structure of regulatory RNAs in a high-throughput manner.⁸⁹

As in the case of target identification, computational structural prediction methods have been important to complement structural probing data. Among the most accurate algorithms are the stochastic methods that can utilize structural data in their algorithms,⁹⁰ the homology/phylogenetic comparative analysis⁹⁹ and thermodynamics-based algorithms.¹⁰³

The need to move toward structure–function relations

An important aspect of being able to fully predict how a synthetic RNA will function is the ability to correlate structural trends to functional capabilities. An interesting recent example of a structure–function study for gene silencing by sRNAs demonstrated a quantitative correlation between gene repression and the predicted free energy of the sRNA–gene target complex.¹⁰⁴

Although there is an arsenal of structure prediction tools available for non-coding RNA computational analysis that have been specifically designed to analyze short non-coding RNAs (reviewed in ref. 105), they have been almost exclusively applied to eukaryotic short non-coding RNAs (e.g., micro RNAs). To the best of our knowledge, there are neither structural prediction algorithms evaluated in their performance exclusively on bacterial regulatory RNAs nor methods specifically tailored to predict bacterial regulatory RNA structures. We can only reasonably expect the current available algorithms, to predict short RNA structure with improved accuracy. In addition, despite the prevalence of hairpins in close proximity in bacterial small regulatory RNAs, to the best of our knowledge, the role of pseudoknots has not yet been explored in these molecules. Pseudoknots are RNA structural motifs in which two stem loops base pair with each other and are currently a challenge for prediction algorithms.

Direct implementation of computational and predictive tools into the creation of synthetic elements

A fascinating recent example of the integration of mechanistic modeling and kinetic RNA modeling simulations to design regulatory RNA elements with desired characteristics has been the dynamic control of glmS riboswitch-like gene expression regulators.⁶⁰ Using a rational design approach, a method to construct RNA-regulated genetic devices was designed based on both biochemical and biophysical predictions. Specifically, static ribozyme- and dynamic aptazyme- regulated genetic devices with quantitatively predictable functions were engineered with the aid of coupled coarse-grained mechanistic and molecular folding simulations. The success of this innovative approach was confirmed by the extraordinary statistical fit that was observed between theoretical predictions and experimental observations. The framework developed in this work represents an important advancement towards applications of synthetic RNA device design. The development of computational methods represents an enormous opportunity in the field of RNA Synthetic Biology.

Unexplored diversity of natural riboswitches in living systems

Identification and characterization of additional RNA scaffolds would largely expand the current repertoire of tools to design novel riboswitches or to further engineer natural ones for new functionalities.²³ However, our understanding of these molecules is based on a small fraction of all the hypothesized naturally existing ones.¹⁰⁶ Several strategies have been used to diversify the catalog of existing riboswitches. For instance, covariance search methods have been applied to predict the presence of riboswitches in a wide range of organisms. Although these studies have been instructive in revealing mechanistic variations in gene control between microbial groups (as predicted by different structural and molecular organizational features¹⁰⁶), the identification of putative riboswitches is based on the limited set of known riboswitch characteristics in a few organisms like E. coli and B. subtilis.

Other creative approaches have attempted to assess the abundance and diversity of riboswitches in microbial communities by performing large-scale metagenomic searches via massive computational and experimental explorations;¹⁰⁷ these studies have also projected the existence of hundreds of different riboswitches (~300) that await validation and characterization. However, a major drawback in the analysis of large-scale metagenomic sequence data sets is the isolation of functional elements, given that these RNAs tend to: (1) be short (often mistaken for chemically degraded RNAs), (2) lack open reading frames, and (3) sometimes evolve rapidly at the sequence levels while still conserving structure that is integral to their function.^107,108

Need for novel in vitro RNA stabilization methods

Although new sensing capabilities of natural and unnatural ligands have been successfully engineered into RNAs via in vitro and in vivo molecular evolution approaches,^109-111 the ability to use these molecules by incorporating them into more complex systems and devices represents a bigger practical challenge.²³ All potential RNA Synthetic Biology applications require reproducible and controlled molecular stability, especially when activity is required outside the native cellular host and in the context of other biological and inorganic materials. Examples of these stringent applications are the targeted delivery of therapeutic RNAs (e.g., siRNAs) from a polymeric matrix;¹¹² these same delivery tools are now being explored for introduction of other RNA synthetic elements in cells. Currently, we lack knowledge on how to mimic the stabilizing effects of protein surfaces since mechanistic aspects of the use of proteins to protect RNAs in biological systems remain largely unexplored. Several strategies have been used to address RNA instability. For instance, molecular inhibitors of cellular nucleases (e.g., RNase E) have been designed to prevent molecular degradation.¹¹³ Still, other creative methods have focused on optimizing intrinsic RNA stability by engineering special secondary motifs and nucleotides to change the chemistry and structure of RNAs themselves in a way that mimics natural post-transcriptional modifications.¹¹⁴ However, one major drawback of directly manipulating RNAs is that only a very small region of the molecule can be targeted (only a few angstroms in size) without loss of desired RNA function. As such, efficient methods for improving RNA stability are another area that awaits further development.

The future of artificial bacterial RNA regulators and their applications

Given the myriad of possible modular combinations using the known RNA parts, and the ones still to be discovered, the possibilities of constructing de novo RNA devices based on regulatory RNAs depend only on our imagination. As we continue to gather fundamental understanding on the principles underlying RNA structure, interactions, and function, the expectations are that we would be able to rationally engineer RNA devices at the smallest possible resolution: single nucleotide (or even single atom). We are already starting to take the first steps toward this as different research groups have already rationally designed riboregulators based on RNA motifs e.g., YUNR consensus sequence⁷³ and kissing–loop interactions^74,75 (Fig. 6).

Overall, the main challenge in the field of synthetic RNA biology is to turn all the special tricks that RNAs naturally perform into useful applications. A further challenge is to be able to synthesize de novo RNA molecules at the atomic level. Although the field has been rapidly moving in this direction, critics still await for major breakthroughs that will yield “useful things” that can directly impact human lives. So far, a variety of applications that stem from “lower hanging fruits” to truly spectacular innovations have been envisioned. Recent work also indicates that additional potential applications could be on their way. Specifically, it seems highly plausible to see an explosion of direct applications of RNA-based Synthetic Biology in bacteria and other organisms traditionally used for production of biochemical compounds. For instance, in enzyme evolution for metabolic engineering applications, Michener et al. have developed an approach to use metabolite-sensing ribozymes linked to gene reporters to evolve and improve enzymatic activity in yeast in a high-throughput way.¹¹⁵ This could have direct application to the bio manufacturing of a variety of compounds that are already produced in E. coli. In addition, in medicine, naturally occurring E. coli could be further explored as a living sensor seeking disease sites and as a therapeutic agent to treat disease-specific sites. The futuristic and almost surreal concept of a therapeutic bacterium is supported by recent works, including Sinha et al. who engineered E. coli to seek and destroy a herbicide⁵⁸ and by the use of non-RNA-based strategies to search for and control cancer cells using Vibrio fischeri.¹¹⁶

Finally, bacteria show potential for use in the development of RNA devices that can be transferred to other cellular platforms for more relevant applications since many RNA functions are considered portable and as such can be transferred from one kingdom to the other with no considerable additional adaptation.^117,118 Additionally, the highly valuable ability of bacterial sRNAs to modulate multiple pathways in metabolic engineering applications could strengthen the potential of using bacteria as a bio-factory.

Other potential fields of application for synthetic regulatory RNAs in a bacterial context that have been explored using either higher organisms or other non-RNA-based technologies include clinical applications like cell therapy and regenerative medicine, microbiome engineering, functional genomics, drug testing, etc.²² In addition, we predict that other non-clinical applications in fields like biosafety and anti-biofouling (e.g., pipes, boats, implantable devices, surgical instruments) will be in the pipeline soon.

Final Thoughts

In a way that parallels how the discovery of large catalytic RNAs revolutionized our perceived limits of biotechnology and bioengineering, the more recent realization of smaller RNA transcripts as powerful regulators has continued to shake our imagination as to the many potential applications of these regulatory molecules in Synthetic Biology. Yet, one of the main challenges faced by this field is the continual fundamental understanding and characterization of the wealth of mechanistic information contained within these molecules. How can we find new RNA scaffolds that display entirely new functional mechanisms in a high-throughput way? Are there more efficient and creative ways to identify cellular targets and pathways that are directly affected by a particular non-coding RNA? What can evolutionary studies tell us about the need for these mechanisms across species? How can we best capture ways by which these molecules are stabilized inside living systems and transfer that knowledge to inorganic surfaces? How can we design rules to predict optimal combinations of multiple RNA units in tandem to tune sensing capabilities? Or, yet, how do we arrange several RNA motifs (biological parts) to increase the functional complexity of RNA devices? How do we accurately predict the behavior of such complex RNA devices from individual parts? How do we develop reliable techniques to profoundly understand underlying principles of complex RNA devices and their ultimate phenotypic impact? So far, we have only touched the surface of all these questions by focusing on a handful of riboswitches and small RNAs via different approaches: rational design (sometimes computationally aided), molecular evolution, and the use of natural elements found in nature.^21,34 Increased fundamental understanding in this area will continue to broaden our creativity in exploiting the many different types of non-coding RNAs that we suspect are present in nature in a much wider range of impactful applications.

Similarly, more insights into the questions we pose will lead to being able to “catch up” applications of RNA biology to those being currently executed by proteins as these are, overall, much better understood. While other molecules (like proteins) might pose advantageous properties over RNAs in different contexts, the addition of RNA to the toolkit will add one more avenue that by itself (or in combination with other types of molecules) will highly contribute to the ongoing creativity of the synthetic biology field.

10.4161/rna.27102