New insights into riboswitch regulation mechanisms


  • Laurène Bastet,

    1. Département de biologie, Faculté des sciences, Groupe ARN/RNA Group, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1K 2R1.
    Search for more papers by this author
  • Audrey Dubé,

    1. Département de biologie, Faculté des sciences, Groupe ARN/RNA Group, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1K 2R1.
    Search for more papers by this author
  • Eric Massé,

    1. Département de biochimie, Faculté de médecine et des sciences de la santé, Groupe ARN/RNA Group, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1H 5N4.
    Search for more papers by this author
  • Daniel A. Lafontaine

    Corresponding author
    1. Département de biologie, Faculté des sciences, Groupe ARN/RNA Group, Université de Sherbrooke, Sherbrooke, Québec, Canada, J1K 2R1.
      E-mail; Tel. (+1) 819 821 8000, ext 65011; Fax (+1) 819 821 8049.
    Search for more papers by this author

E-mail; Tel. (+1) 819 821 8000, ext 65011; Fax (+1) 819 821 8049.


Riboswitches are genetic elements located in non-coding regions of some messenger RNAs (mRNAs) that are present in all three domains of life. The binding of ligands to riboswitches induces conformational changes in the mRNA molecule, resulting in modulation of gene transcription, or RNA splicing, translation or stability. This mechanism of regulation is particularly widespread in bacteria and allows a direct response to various metabolic changes. A large number of riboswitches have been discovered in the last few years, suggesting the existence of a huge diversity of regulatory ligands and genetic mechanisms of regulation. This review focuses on recent discoveries in riboswitch regulatory mechanisms as well as current outstanding challenges.


A complex network of diverse interplaying pathways allows bacteria to use nutrients from their surroundings to adapt to environmental changes. This outstanding capacity of adaptation requires bacteria to orchestrate signal perception, transmission of information, gene expression regulation and ultimately adjustment of cellular metabolism. Most of the well-characterized bacterial systems responding to external conditions are composed of proteins acting as sensors and/or regulators. However, an increasing number of studies have shown that regulatory non-coding RNA structures, called riboswitches, play a crucial role in bacterial adaptive responses through sensing cellular metabolites.

Riboswitches are composed of two functional domains, the aptamer and the expression platform. The aptamer is the sensor domain of the riboswitch and is very conserved across bacterial species as it is involved in the recognition of the signal. The second domain consists in the expression platform that is determinant for the regulation mechanism. The detection of a biological signal causes the mRNA to undergo a global conformational change that modulates gene expression. Thus, because riboswitches can both sense cellular signals and regulate gene expression, they allow a direct response to metabolic stress.

As aptamer domains are highly conserved in sequence and structure, they have been used to classify riboswitches accordingly to the regulatory signal. Riboswitches can sense cellular metabolites such as amino acids and derivatives [lysine, glycine, S-adenosylmethionine (SAM), S-adenosylhomocysteine (SAH)], carbohydrates [glucosamine 6-phosphate (Glcn6P)], coenzymes [flavin mononucleotide (FMN), thiamin pyrophosphate (TPP), coenzyme B12], nucleobases and their derivatives [adenine, guanine, cyclic di-GMP, pre-queuosine (preQ1)] (Barrick and Breaker, 2007; Sudarsan et al., 2008), metal ions (molybdenum cofactor and magnesium ions) (Regulski et al., 2008; Spinelli et al., 2008), tRNA (Henkin, 2009) and physico-chemical parameters such as temperature and pH (Morita et al., 1999; Nechooshtan et al., 2009).

Bacterial riboswitches control biological processes at various regulatory levels, such as transcription and translation (Waters and Storz, 2009). In the case of transcriptionally acting riboswitches, signal detection induces the formation of a Rho-independent terminator that is characterized by a GC-rich stem followed by a poly-U tail. The formation of the terminator destabilizes the RNA polymerase and produces an arrest of transcription elongation. Translationally controlling riboswitches modulate the ribosome access to the Shine–Dalgarno or AUG start codon sequences in a stem-loop structure. In addition to these two mechanisms, the glmS riboswitch has been recently shown to perform self-cleavage when bound to glucosamine 6-phosphate, leading to rapid mRNA degradation (Collins et al., 2007). Riboswitches can also be combined to allow more complexity in gene regulation response. For example, the Bacillus subtilis gcvT gene is regulated by a tandem riboswitch composed of two glycine aptamers followed by one expression platform (Mandal et al., 2004). Because glycine binds cooperatively to the riboswitch, it enables gene regulation to exhibit a more ‘digital’ character by allowing efficient control in response to slight changes in glycine concentration. In other cases, two different riboswitches are associated with the regulation of the same downstream gene. As such, riboswitches responding to SAM and AdoCbl are located adjacently to each other in Bacillus clausii (Sudarsan et al., 2006), where each riboswitch controls its own expression platform. Another example of combined regulation mechanisms involves the Mg2+-sensing riboswitch of Salmonella thyphimurium controlling mgtA expression. This riboswitch acts at the transcriptional level as well as mRNA stability (Spinelli et al., 2008). In addition, the mgtA gene expression is dependent on proteins PhoP/PhoQ, which is a two-component system responding to magnesium concentration (Cromie and Groisman, 2010).

Bioinformatic studies and high throughput approaches have predicted the existence of new riboswitches in bacterial genomes (Barrick et al., 2004; Weinberg et al., 2007; Irnov and Winkler, 2010). Various riboswitch regulatory mechanisms have also been recently discovered providing evidence about the complexity and diversity of RNA-based mechanisms. This review will focus on novel riboswitch-dependent gene regulatory mechanisms. The first example discussed here describes how physico-chemical changes can be harnessed by riboswitches to regulate gene expression (Narberhaus et al., 2006; Nechooshtan et al., 2009). The second regulation mechanism concerns the stability and function of riboswitches generated by transcriptional arrest or RNase cleavages (Shahbabian et al., 2009). These recent discoveries open new research hypotheses and perspectives related to riboswitch regulation mechanisms that will be discussed. Future directions and experimental approaches will be proposed as a way to address open questions in the riboswitch field.

Riboswitches sensing physico-chemical changes

Most characterized riboswitches regulate gene expression through the binding of a specific ligand. However, various RNA elements respond directly to changes at the physico-chemical level. One of such riboswitches is a temperature-sensing riboswitch, or thermosensor, which is sensitive to temperature variations (Narberhaus et al., 2006). Generally, thermosensors block translation at low temperature by sequestering the Shine–Dalgarno and initiation codon sequences into a stem-loop structure. Following a rise in temperature, the riboswitch structure is denatured and translation signals are then accessible to ribosomes to allow ribosome binding and translation initiation. An interesting issue concerns how physico-chemical parameters can induce specific RNA conformational changes. Narberhaus and others have suggested that the folding of thermosensors is dependent on a temperature gradient, consistent with a model in which the RNA helical stability is important for regulation (Nocker et al., 2001; Johansson et al., 2002; Narberhaus et al., 2006; Rinnenthal et al., 2010). It was proposed that the melting of RNA near the Shine–Dalgarno sequence would increase the accessibility to ribosome binding. However, according to another study, there was no significant difference between thermosensors and random RNA sequences by comparing their melting properties (Shah and Gilchrist, 2010). Thus, specific criteria that make these RNA structures more sensitive to temperature still need to be established. As Shine–Dalgarno sequences could in principle be temperature-sensitive in many mRNAs, Shah and Gilchrist have suggested that the structural destabilization observed for RNA thermometers was probably not the only criteria defining these regulation mechanisms (Shah and Gilchrist, 2010).

Recent work has unveiled a new thermosensor that adopts an active structure following a cold shock treatment. The structural change allows the translation of cspA, which encodes for a chaperon protein (Giuliodori et al., 2010). The expression of cspA is favoured at cold temperatures as its function is to bind single-stranded RNAs in order to prevent the stabilization of some RNA structures at low temperatures (Jiang et al., 1997). This is the first riboswitch known to allow translation under cold conditions.

Recently, another riboswitch responding to physico-chemical conditions has been reported (Nechooshtan et al., 2009) (Fig. 1). This RNA element is located in the 5′UTR of the uncharacterized alx gene. Transcriptomic data have shown that the alx gene is upregulated under alkaline conditions (Bingham et al., 1990). Subsequently, Nechooshtan and coworkers have demonstrated by way of translational fusions and in vitro experiments that this RNA element reacted to alkaline pH and regulated alx expression at the translational level (Nechooshtan et al., 2009). Interestingly this riboswitch must be transcribed de novo to allow mRNA translation when cellular conditions become permissive. Moreover, it was clearly demonstrated that pH-dependent translational regulation relied on transcriptional pausing (Nechooshtan et al., 2009). Previously, only transcriptionally acting riboswitches were known to be dependent on transcription kinetics (Wickiser et al., 2005; Lemay et al., 2011). Indeed, in the case of the FMN riboswitch, RNA polymerase pausing sites allow time for the aptamer to bind to the ligand before the genetic decision is made. In the case of the pH-responsive riboswitch, two pausing sites have been identified in which pause lifetimes are increased under alkaline condition (Nechooshtan et al., 2009). Thus, transcriptional pausing promotes the active conformational state and allows the translation of the alx gene.

Figure 1.

Model illustrating the mechanism of a translationally acting pH-dependent riboswitch. At neutral pH, transcription of the alx mRNA leads to the formation of the stems A, B, C and D. In these circumstances, the D stem prevents ribosome access to Shine–Dalgarno (gray box) and AUG start codon (black box) sequences. However, under alkaline conditions, RNAp stops at two different pausing sites (asterisks), leading to the formation of the C/D stem-loop. In this alternative structural arrangement, the alx mRNA releases translation signals thus allowing ribosome translation. RNAp, RNA polymerase.

Riboswitch RNAs: from cis-acting riboswitch to trans-acting small RNA

Transcriptionally acting riboswitches prevent RNA polymerase elongation via the formation of a transcription terminator. These premature transcriptional arrests generate short RNAs corresponding to the riboswitch sequence (Fig. 2). Aptamer fragments can also be produced by RNase cleavages of transcriptionally acting riboswitches, frequently leading to mRNA destabilization (Spinelli et al., 2008; Shahbabian et al., 2009). Currently, the exact mechanisms by which riboswitch RNA fragments are generated are not well characterized and need further investigation. Nevertheless, an important question prevails: what are the roles of these fragments?

Figure 2.

Schematic representation of the generation and the fate of riboswitch RNAs. Riboswitch fragments can originate from RNase cleavage or transcriptional arrest from Rho-independent or Rho-dependent signals. Three possible outcomes can be considered for riboswitch RNAs: (i) to regulate gene expression in trans by base pairing to another mRNA, (ii) to decrease the intracellular concentration of a particular ligand by titration and (iii) to allow ligand recycling by quick RNA degradation. RNAp, RNA polymerase; ORF, open reading frame.

Recently, Shahbabian and coworkers have reported that short RNAs generated from the transcriptionally acting SAM riboswitches of B. subtilis are degraded post-transcriptionally (Shahbabian et al., 2009). Indeed, the analysis of various RNase mutants revealed that a novel endoribonuclease, RNase Y, initiates the cleavage of the riboswitch mRNA species, which is then degraded by the 3′-5′ exoribonucleases PNPase and RNase R. The authors also found that the RNase Y activity was dependent on the presence of the ligand. The rate of cleavage increases in presence of SAM, suggesting that cleavage sites are more exposed when the riboswitch is bound to SAM. Furthermore, the RNase Y-mediated degradation decreases the mRNA stability, which is a mechanism that has been previously described for the transcriptionally acting magnesium-sensing riboswitch of S. thyphimurium (Spinelli et al., 2008). In the case of the magnesium-sensing riboswitch, magnesium binding enhances the cleavage of the aptamer by RNase E, causing the fast degradation of the mRNA. In both SAM- and Mg2+-sensing riboswitches, the control of the mRNA stability seems to act jointly with transcriptional regulation mechanisms to efficiently modulate gene expression. However, unlike the B. subtilis SAM riboswitch, the magnesium riboswitch is detectable both by qRT-PCR and Northern blot experiments (Kawano et al., 2005; Spinelli et al., 2008). These observations suggest intriguing mechanistic differences in riboswitch regulation given that some riboswitch RNAs are abundant in the cell while others appear to be quickly degraded.

In principle, bacteria could benefit from maintaining a high level of riboswitch RNAs. For instance, a high aptamer concentration could allow titration of the metabolite (Fig. 2), effectively leading to a decrease of its cellular free concentration and to the maintenance of homeostasis (Kawano et al., 2005). Recently, an unexpected role for riboswitches has been uncovered, providing a rationale for their cellular stability (Loh et al., 2009). Loh and coworkers showed that two SAM riboswitches from Listeria monocytogenes (SreA and SreB) act in trans similarly to small regulatory RNAs, in addition to their cis-regulatory function (Fig. 2). Small RNAs are known to regulate gene expression through base pairing to the target mRNA by modulating translation and mRNA degradation (Caron et al., 2010). In the case of SAM riboswitches, it was shown that SreA and SreB generated by transcriptional arrest interacted with several mRNAs (Loh et al., 2009). Among the cellular targets of SreA and SreB, prfA mRNA is known to be involved in the positive control of virulence genes. During infection with L. monocytogenes, PrfA expression is upregulated by a thermosensor riboswitch that releases the Shine–Dalgarno and the initiation codon at 37°C (Johansson et al., 2002). However, because the SreA riboswitch can also directly pair to the thermosensor, it represses the expression of PrfA during the infection process. Moreover, because the interaction site is located approximately 80 nucleotides upstream of the translation initiation signal, it is still unclear how SreA downregulates the expression of prfA. The trans-acting role for a riboswitch emphasizes the importance of RNA in the regulation of gene expression, and considering the abundance of riboswitches in the cell, it is highly probable that other riboswitches act in trans as a way to cross-talk between multiple metabolic pathways.


Future challenges will reside in the findings and validation of new riboswitches to gain a more complete understanding of RNA-based gene regulation mechanisms. Based on recent discoveries in the riboswitch field, the following guidelines propose an up-to-date perspective regarding concepts that are deemed to be important for validation of new riboswitch candidates. These concepts are based on the findings of conserved aptamer structures, the identification of corresponding cellular metabolites and the in vivo validation of new riboswitches using gene expression reporter systems.

Identification of new riboswitch candidates

Recent advances in the riboswitch field suggest the existence of a huge variety of aptamers, and consequently, of cellular signals detected by RNA molecules. For example, in silico approaches predict the presence of novel riboswitch candidates in various bacterial genomes (Barrick et al., 2004; Kazanov et al., 2007; Weinberg et al., 2007). These studies are based on specific riboswitch characteristics such as the conservation of structures and nucleotide sequences of the aptamer domain. Additional parameters consist in the localization of the element in the 5′UTR of mRNA as well as the presence of a singular structure in the 3′ extremity of the motif that could correspond to a known expression platform. According to those predetermined characteristics, six putative riboswitches have been previously predicted: glmS, gcvT, ykoK, ydaO/yuaA, ykkC/yxkD and yybp/ykoY (Barrick et al., 2004). Three of these motifs have been shown to be riboswitch regulatory elements. The ykoK motif senses Mg2+ concentrations, the glmS riboswitch detects the glucosamine 6-phosphate and the yybp/ykoY motif, also named SraF, responds to alkaline pH. The fact that the presence of some of these putative riboswitches has been experimentally confirmed supports the validity of bioinformatics to uncover novel riboswitches.

Moreover, in silico data can be consolidated by a more global experimental approach based on sequencing all RNA retrieved from bacteria. Indeed, this technique has been recently used to identify regulatory RNA in B. subtilis (Irnov et al., 2010). With this method, new small regulatory RNAs and long 5′UTR in mRNAs have been discovered. Given that the size of mRNA 5′UTR regions is usually less than 35 nucleotides, it suggests that long 5′UTR might be biologically relevant regulatory RNAs involved in the regulation of downstream genes. The combination of in silico analysis and high throughput sequencing would allow a more efficient detection of novel riboswitch candidates.

As a way to validate new riboswitch candidates, putative ligands inducing RNA conformational changes and gene regulation ought to be identified. Most of the time, an obvious candidate can readily be deduced from the regulated genes. Frequently, riboswitch-controlled genes encode proteins involved in ligand biosynthetic pathways or transporters. However, in the case of some riboswitches, the identification of the ligand is less obvious. For example, the ydaO motif possesses all riboswitch features (Block et al., 2010) and yet, downstream genes are too diversified or not yet characterized enough to elucidate the identity of the regulating ligand. Transcriptomic data can be used to predict possible ligands involved in riboswitch regulation. The role of pH in the riboswitch regulation of alx has been determined using microarray data (Nechooshtan et al., 2009). Thus, experimental conditions for which the expression of the riboswitch downstream gene is affected are crucial to trace back the regulating ligand. The identification of new ligands can also benefit from high throughput approaches to rapidly assess different families of molecules in the modulation of studied riboswitches (Blount et al., 2006; Mayer and Famulok, 2006).

A cell-based approach can also be used to determine riboswitch-regulating metabolites. For example, by employing a construct comprising the riboswitch fused to a reporter gene, it is possible to search for cellular conditions that affect riboswitch modulation of reporter gene expression. If such a condition is identified, it becomes then possible to design biochemical protocols to isolate cellular metabolite(s) that are responsible for gene induction. A metabolomic approach could be used to determine metabolites produced from growth conditions in which the riboswitch modulation is observed (Winder et al., 2008; Tremaroli et al., 2009; Yuan et al., 2009). Ideally, metabolites that are more abundant in the predetermined conditions should be tested in vitro using approaches such as in-line probing for their capacity to bind the putative riboswitch (Soukup and Breaker, 1999). The identification of a regulating metabolite is crucial to validate riboswitch candidates and to identify mRNA structural changes involved in gene regulation.

From new riboswitches to new regulation mechanisms

Initially, riboswitches were regarded to regulate gene expression only at transcriptional and translational levels. However, several riboswitches have recently been described to modulate gene expression at other levels, such as mRNA stability, and by using more complex mechanisms such as tandem aptamers or trans-acting regulation (Serganov and Patel, 2007).

Interestingly, the research on known or putative riboswitch classes in different organisms has revealed a huge variety of structures that expression platform can adopt. Visual inspection of various expression platforms does not readily allow to infer a particular regulation mechanism (R. Tremblay and D.A. Lafontaine, unpubl. obs.). This finding suggests that the repertoire of riboswitch regulation mechanisms may actually be broader than currently known. In vivo studies based on mutational analysis, Northern blot hybridization, transcriptional and translational fusions using a reporter gene will be necessary to address these new mechanisms. In addition, regulation mechanisms can also exploit truncated riboswitch molecules generated from RNase cleavages or transcriptional arrests (Spinelli et al., 2008; Loh et al., 2009; Shahbabian et al., 2009). It will be particularly interesting to explore thoroughly these mechanisms and to determine whether riboswitch fragments are involved in other biological functions.


L.B. holds a Bourse doctorale triennale from the Université de Sherbrooke. This work was funded by operating grants MOP69005 and MOP82877 to E.M. and D.A.L., respectively, from the Canadian Institute for Health Research (CIHR). E.M. and D.A.L. are Chercheur-boursier Junior 2 from the Fonds de la Recherche en Santé du Québec (FRSQ). D.A.L. is also a New Investigator scholar from the CIHR.