Riboswitch function: flipping the switch or tuning the dimmer?

Abstract

Riboswitches are structured mRNA elements involved in gene regulation that respond to the intracellular concentration of specific small molecules. Binding of their cognate ligand is thought to elicit a global conformational change of the riboswitch, in addition to modulating the fine structure of the binding site. X-ray crystallography has produced detailed descriptions of the three-dimensional structures of the ligand-bound conformations of several riboswitches. We have employed small-angle X-ray scattering (SAXS) to generate low-resolution reconstructions of the ligand-free states of the ligand-binding domains of riboswitches that respond to thiamine pyrophosphate (TPP), and cyclic diguanylate (c-di-GMP), a bacterial second messenger. Comparison of the SAXS reconstructions with the crystal structures of these two riboswitches demonstrates that the RNAs undergo dramatic ligand-induced global conformational changes. However, this is not an universal feature of riboswitches. SAXS analysis of the solution behavior of several other riboswitch ligand-binding domains demonstrates a broad spectrum of conformational switching behaviors, ranging from the unambiguous switching of the TPP and c-di-GMP riboswitches to complete lack of switching for the flavin mononucleotide (FMN) riboswitch. Moreover, the switching behavior varies between examples of the same riboswitch from different organisms. The range of observed behaviors suggests that in response to the evolutionary need for precise genetic regulation, riboswitches may be tuned to function more as dimmers or rheostats than binary on/off switches.

Introduction

Riboswitches are structured domains of the untranslated regions and introns of mRNAs that directly recognize small molecule metabolites and second messengers and regulate gene expression in cis. While some riboswitches execute genetic control through ligand-induced self-cleavage^{1, 2} or hybridization in trans,³ most riboswitches are thought to undergo a conformational change upon binding their cognate effector molecule, which subsequently translates into a genetic on/off switch.

Riboswitches can be thought of as being comprised of two distinct, separable domains, one responsible for small molecule binding (the 'aptamer domain') and another for transducing small molecule binding into a genetic regulatory signal (the 'expression platform').⁴ The former exhibits the majority of the sequence conservation in the riboswitch and it alone is responsible for recognition of the cellular small molecule. Because of this, many studies have focused on the aptamer domain in an effort to understand riboswitch function, including the role of conformational switching induced by cognate ligand binding. Three-dimensional structures have been reported for 11 of the approximately 20 different classes riboswitch aptamer domains discovered to date, all in their ligand-bound conformations (reviewed in refs. 5–7).

Modes of small molecule recognition

Crystallographic analyses in our laboratory and others has revealed the molecular details of small-molecule recognition by the TPP and c-di-GMP riboswitch aptamer domains, which are ~80 and ~90 nucleotides long, respectively.^8–12 The ligand-bound conformations of these aptamer domains are globally similar, despite differences in the size, shape, and overall charge of their respective cognate small molecules (Figure 1). Both RNAs adopt h-shaped folds comprised of two parallel helical elements connected by a three-way junction to a helical 'handle' containing the 5'- and 3'- termini of the aptamer domain. The parallel helices contain irregular helical regions and loops that participate in long-range (tertiary) interactions important for stabilizing the overall fold of the RNA domain.

Figure 1.

Secondary and schematic three-dimensional structures²⁸ of the (A) TPP and (B) c-di-GMP riboswitch aptamer domains. The parallel helices involved in long-range interactions are colored red and blue. Regions involved in ligand binding are colored gold and the ligands are colored purple. The grey region atop of the c-di-GMP aptamer domain comprises the U1A protein and its binding site, introduced to facilitate crystallization.

The most dramatic difference between these two riboswitch aptamer domains is in their modes of small molecule recognition. In the TPP aptamer domain, the central, irregular regions of each of the two parallel helices directly participate in metabolite binding. One folds into a structure resembling the T-loop of tRNA and recognizes the pyrimidine moiety of TPP, while the other contacts the pyrophosphate of TPP, which is bound as a Mg²⁺ chelate. In contrast, the analogous parallel helices of the c-di-GMP aptamer domain utilize their non-A-form central regions for presentation of an interhelical Watson-Crick base pair, rather than for cognate small-molecule binding. In the c-di-GMP aptamer domain, small molecule recognition is achieved within the three-helix junction. This junction consists of 14 nt and binds to both nucleobases of c-di-GMP molecule through a combination of base pairing and stacking interactions. The three-helix junction of the TPP aptamer domain is comprised of only 7 nt, and makes no interactions with the metabolite.

The specifics of the ligand recognition strategies of these aptamer domains could not have been deduced a priori from the molecular properties of their respective cognate small molecules. For instance, that the c-di-GMP would adopt a ‘sandwiched’ conformation in complex with the aptamer domain was not necessarily expected, as some proteins are known to recognize this second messenger in an extended conformation, with the nucleobases splayed out in opposite directions.¹³ In light of the TPP aptamer domain structure, it might have seemed plausible for the c-di-GMP to be recognized in such an extended conformation, bridging the two parallel helices of the RNA. Alternatively, given the 2-fold symmetry of c-di-GMP, a dimer of aptamers could have recognized the second messenger. Indeed, many c-di-GMP binding proteins are symmetric oligomers.^{14, 15}

Mode of recognition predicts nature of global response?

Can riboswitch aptamer domains be classified based on crystal structures of their ligand-bound states, and can such a classification predict the global structural response of the domains to ligand binding? Based on the six structures available at the time, Montange and Batey¹⁶ proposed that riboswitches could be classified into two distinct families. The aptamer domains of Type I riboswitches bind to their respective cognate small molecules at the junction of several helices or at a pseudoknot (e.g. the purine¹⁷ or the glmS¹⁸ riboswitches, respectively) while Type II aptamer domains utilize structural elements distant in secondary structure to bind their ligands (e.g. the TPP riboswitch). The two types can be described as ‘junction’ vs. ‘other’ modes of ligand recognition, because the aptamer domains that do not bind their metabolite in a multi-helical junction (or a pseudoknot) use a variety of distinct peripheral structural elements to achieve specific binding. Furthermore, Montange and Batey¹⁶ suggested that this structural classification would predict the global responses to small molecule binding of the aptamer domain. Specifically, because Type I riboswitches recognize their ligands at a central helical junction, their global architectures are predicted to be preformed and ligand binding is proposed to induce only local rearrangements of the binding pocket. In contrast, Type II riboswitches require distal structural elements to coalesce to form their ligand binding pocket; therefore, they must undergo a global structural rearrangement upon ligand binding.

Given their structures in complex with their respective cognate ligands, the c-di-GMP riboswitch would be classified as a Type I or ‘junction’ riboswitch, whereas the TPP riboswitch would be classified as a Type II, or ‘other’. It follows that the Type I c-di-GMP riboswitch aptamer domain would be prefolded and not undergo global structural rearrangements upon binding the second messenger, while the Type II TPP riboswitch aptamer would be expected to undergo a global rearrangement upon binding its cognate metabolite. Based on the structure of the c-di-GMP riboswitch aptamer domain and in-line probing data, Strobel and colleagues¹² predicted this RNA would be prefolded. In the case of the TPP riboswitch, Micura and colleagues¹⁹ inferred from fluorescence spectroscopic studies of aptamer domain constructs with 2-aminopurine substitutions that the RNA undergoes large-scale folding in response to metabolite binding. Calorimetric analysis of TPP binding by the latter riboswitch is consistent with this inference.²⁰

The c-di-GMP and TPP riboswitches in solution

We employed small-angle X-ray scattering (SAXS) to investigate of the nature of the global ligand-induced structural responses of the c-di-GMP and TPP riboswitch aptamer domains. SAXS is a powerful tool for examining conformations of macromolecules in solution and has been used to monitor RNA folding for over a decade.²¹ From a SAXS experiment, the radius of gyration (R_g) and maximum dimension of the molecule are obtained, providing global measures of the shape of a macromolecule in solution. From a practical standpoint, one advantage of SAXS is that a variety of solution conditions (e.g. a riboswitch without and with saturating concentrations of its cognate ligand) can be readily probed.²² Since the mechanism of riboswitch function may involve a global ligand binding-induced conformational switch, SAXS is an appropriate experimental technique.

We examined the conformation of the TPP and c-di-GMP riboswitch aptamer domains under near-physiologic²³ Mg²⁺ concentrations (1.5 – 2.5 mM) and under highly stabilizing Mg²⁺ conditions (10 mM), both in the absence and presence of their cognate ligands. In 2.5 mM Mg²⁺, the TPP aptamer domain undergoes considerable compaction upon metabolite binding as indicated by the change in R_g (ΔR_g) of 3.8 Å (Table 1). The ΔR_g is smaller (2.3 Å) in the presence of 10 mM Mg²⁺, indicating that highly stabilizing Mg²⁺ is capable of inducing a partially pre-organized conformation relative to physiologic Mg²⁺ (Table 1). Under all Mg²⁺ concentrations examined, this riboswitch aptamer domain undergoes metabolite-induced compaction, albeit to varying degrees (Table 1 and ref. 24) Ab initio reconstructions based on the SAXS data provide low-resolution views of the conformations of this RNA under various conditions (Figure 2).^{24, 25} Comparison of the free and TPP-bound reconstructions indicate that the helical elements involved in metabolite recognition are not pre-arranged in a side-by-side orientation prior to binding TPP (Figure 2).

Table 1.

R_g values in varying Mg²⁺ concentrations for TPP and c-di-GMP aptamer domains.

[Mg2+], mM	Rg, ±1 Å TPP aptamer				Rg, ±1 Å c-di-GMP aptamer
	E. colia		A. thalianab		V. choleraec		T. crunogenad
	free	bound	free	bound	free	bound	free	bound
2.5	25.7	21.9	n.d.	n.d.	28.5	23.9	28.0	25.2
10	24.4	22.1	28	22	25.1	23.9	n.d.	n.d.

^aData taken from ref. 24

^bData taken from ref. 25

^cData taken from ref. 11

^dExperimental conditions for the T. crunogena aptamer domain were as in ref.11

Figure 2.

Low-resolution SAXS reconstructions of the TPP and c-di-GMP riboswitches in the ‘bound’ and ‘free’ conformations. (A) The low-resolution reconstruction of the TPP ‘bound’ aptamer domain is colored as in Figure 1. The location of the red helix in the ‘free’ reconstruction is assigned based on experiments described in ref. 24. The assignment of regions corresponding to the blue and grey helices is somewhat ambiguous.²⁵ (B) The ‘bound’ and ‘free’ reconstructions of the c-di-GMP aptamer domain are colored according to Figure 1 and assigned as described in ref. 11.

SAXS investigation of the c-di-GMP riboswitch aptamer domain indicates that it also undergoes a pronounced global conformational change upon binding its cognate ligand.¹¹ The observed ΔR_g between the free and c-di-GMP-bound conformations in 2.5 mM Mg²⁺ is 4.6 Å, a change even larger than that observed for the TPP riboswitch aptamer domain (Table 1). The magnitude of this ligand binding-induced conformational change is surprising given the prediction of a pre-formed structure that undergoes only local changes upon c-di-GMP binding for this Type I riboswitch.¹⁶ Low-resolution reconstructions of the aptamer domain in the free and bound states are dramatically different from each other (Figure 2). The SAXS reconstruction of the ligand-bound state is in excellent agreement with the cocrystal structure. The reconstruction of the free state of the aptamer domain has two arms whose dimensions are strikingly similar to those of helices P1b and P2. This suggests that in the absence of c-di-GMP, these helices are not oriented side-by-side, but are splayed out in opposite directions, in a manner reminiscent of the ligand-free state of the TPP riboswitch. Back-calculation of scattering profiles from such a model of the ligand-free c-di-GMP riboswitch, as well as nuclease probing experiments, support this conclusion.¹¹

The cocrystal structure and the SAXS reconstructions were obtained with an RNA construct whose sequence corresponds to that of a c-di-GMP riboswitch from Vibrio cholerae.¹¹ The unexpectedly large ligand binding-induced global conformational change of this RNA is probably not an anomaly of this specific riboswitch sequence, because SAXS analysis of the aptamer domain of a second c-di-GMP riboswitch (from Thiomicrospira crunogena) also reveals a ligand-binding induced compaction (Table 1). In 2.5 mM Mg²⁺ and the absence of c-di-GMP, this RNA exhibits an R_g = 28.0 Å. Upon the addition of saturating c-di-GMP, the aptamer domain compacts to an R_g = 25.2 Å.

Diversity of riboswitch responses to ligand

It appears that the structural responses of riboswitch aptamer domains to ligand binding are far more nuanced than implied by the binary classification of Montange and Batey¹⁶. For example, independently from our studies of an Escherichia coli TPP riboswitch, Doniach and colleagues examined by SAXS the aptamer domain of a TPP riboswitch from Arabidopsis thaliana.²⁵ The TPP riboswitch aptamer domains from the two organisms have contrasting behaviors. In highly stabilizing Mg²⁺ conditions (10 mM) and in the absence of metabolite, the TPP riboswitch aptamer domain from E. coli exhibits an R_g = 24.4 Å. In contrast, a much more expanded conformation (R_g = 28 Å) is observed for the RNA from A. thaliana under the same conditions.²⁵ In the presence of 10 mM Mg²⁺ and saturating concentrations of TPP, the R_g for both of these aptamers compacts to 22 Å, which is consistent with the observation that the cocrystal structures of the two RNAs are very similar.⁹ The distinctly different metabolite-free conformations of these two TPP riboswitch aptamer domains demonstrate that small molecule-driven structural responses are not comparable even between orthologous riboswitches from different organisms.

The idiosyncratically tuned nature of the global metabolite binding-induced responses of riboswitch aptamer domains extends across riboswitch classes as well. We employed SAXS to investigate riboswitch aptamer domains that bind S-adenosyl methionine (SAM-I), lysine, and FMN.²⁴ We found that the SAM-I aptamer domain behaves similarly to the TPP riboswitch aptamer domain in 2.5 mM Mg²⁺ in that it displays a ΔR_g of ~ 3 Å between its free and bound conformations. However, in the absence of SAM, more stabilizing Mg²⁺ concentrations (10 mM) compact the RNA to within 1 Å of its bound conformation, demonstrating a global response that is highly dependent on the environmental conditions. The global conformation of the lysine riboswitch aptamer domain is insensitive to the presence of its cognate ligand under all Mg²⁺ concentrations examined. Instead, a significant conformational change is induced by varying Mg²⁺ concentrations. Finally, the FMN riboswitch aptamer domain is globally insensitive to metabolite binding as well as Mg²⁺ concentrations: this RNA is already maximally compact in 1.5 mM Mg²⁺ in the absence of metabolite. Altogether, our SAXS experiments show that the global conformational response of riboswitch aptamer domains does not neatly follow the classification proposed by Montange and Batey.¹⁶ At least one Type I aptamer domain (c-di-GMP) undergoes a global switch, and Type II aptamer domains (SAM, TPP) display a range of behaviors in response to ligand binding and solution conditions.²⁴

Conclusion: riboswitches, digital or analog?

The diversity in the solution behavior of the aptamer domains revealed by SAXS analysis suggests that riboswitch response to ligand binding is better described as analog than digital. The degree of metabolite-induced aptamer switching is likely variably tuned, similar to the precise control offered by a dimmer switch, in contrast to the binary open or closed states of a single pole switch. Analysis of the regulation of gene expression by riboswitches suggest that they rarely function as simple digital “on/off” switches. In vivo and in vitro studies of the transcriptional regulation of gene expression by SAM-I and guanine riboswitches from different loci in the Bacillus subtilis genome revealed a wide range of transcript levels, transcriptional read-through efficiencies, and binding affinities for twelve SAM-I riboswitches²⁶ and five guanine riboswitches.²⁷ These studies suggest that the response of these riboswitches is tuned by sequence changes outside the near-invariant nucleotides directly involved in small molecule recognition. These sequence changes may reflect adaptation of each riboswitch for efficient control of gene expression within its own genomic locus.