Ligand binding by the tandem glycine riboswitch depends on aptamer dimerization but not double ligand occupancy

Abstract

The glycine riboswitch predominantly exists as a tandem structure, with two adjacent, homologous ligand-binding domains (aptamers), followed by a single expression platform. The recent identification of a leader helix, the inclusion of which eliminates cooperativity between the aptamers, has reopened the debate over the purpose of the tandem structure of the glycine riboswitch. An equilibrium dialysis-based assay was combined with binding-site mutations to monitor glycine binding in each ligand-binding site independently to understand the role of each aptamer in glycine binding and riboswitch tertiary interactions. A series of mutations disrupting the dimer interface was used to probe how dimerization impacts ligand binding by the tandem glycine riboswitch. While the wild-type tandem riboswitch binds two glycine equivalents, one for each aptamer, both individual aptamers are capable of binding glycine when the other aptamer is unoccupied. Intriguingly, glycine binding by aptamer-1 is more sensitive to dimerization than glycine binding by aptamer-2 in the context of the tandem riboswitch. However, monomeric aptamer-2 shows dramatically weakened glycine-binding affinity. In addition, dimerization of the two aptamers in trans is dependent on glycine binding in at least one aptamer. We propose a revised model for tandem riboswitch function that is consistent with these results, wherein ligand binding in aptamer-1 is linked to aptamer dimerization and stabilizes the P1 stem of aptamer-2, which controls the expression platform.

INTRODUCTION

Riboswitches are noncoding elements in mRNA that modulate expression of a gene in response to changes in the concentration of a specific small-molecule ligand. The glycine riboswitch, which senses the concentration of the smallest amino acid (Mandal et al. 2004), is a common riboswitch, with at least 350 known instances spread across the bacterial kingdom (Kazanov et al. 2007; Kladwang et al. 2012). Of these, ∼60% are found regulating genes for the glycine cleavage system, primarily gcvT and gcvP (Barrick and Breaker 2007; Kazanov et al. 2007). Another ∼20% of the known glycine riboswitches regulate a sodium/alanine- or glycine-symporter (Mandal et al. 2004). At concentrations in excess of that necessary for protein synthesis, glycine binds to the riboswitch and activates glycine cleavage genes, breaking down glycine into ammonia, a methylene unit in the form of methyl-THF, and NADH (Barrick et al. 2004). In addition to exploiting excess glycine as an energy source, bacteria must regulate concentrations to prevent glycine toxicity. High concentrations of glycine interfere with cell wall biosynthesis, and glycine riboswitches are necessary for Streptomyces griseus populations to grow in high glycine (Tezuka and Ohnishi 2014). The riboswitch system allows organisms to respond quickly to changes in local environment, catabolizing glycine for energy when environmental concentrations get too high, and importing glycine from the environment when concentrations fall too low.

The glycine riboswitch predominantly exists in a tandem architecture, with two adjacent, homologous aptamers joined by a short linker region, followed by a single expression platform (Fig. 1A; Mandal et al. 2004). The two aptamers each bind glycine with micromolar dissociation constants (1–30 µM K_d), and, like many other riboswitches, the glycine riboswitch undergoes conformational compaction upon the addition of glycine (Mandal et al. 2004; Lipfert et al. 2007; Kwon and Strobel 2008; Baird and Ferré-D'Amaré 2013; Zhang et al. 2014). It is the only known riboswitch where two aptamer domains control a single expression platform (Mandal et al. 2004; Breaker 2011). Because the tandem architecture, rather than a simpler single-aptamer riboswitch, has been conserved against evolutionary drift, it is expected to provide some benefit, possibly for ligand-binding affinity, kinetic response time, or complex genetic control.

FIGURE 1.

Secondary and tertiary structures of the tandem glycine riboswitch. (A) Secondary structure showing the tandem glycine aptamers, each binding one equivalent of glycine. Here, a transcriptional on-switch is depicted. The recently identified leader helix is boxed, and the P0 helix is shown. (B) Secondary and tertiary structures of the tandem glycine riboswitch from Fusobacterium nucleatum (Fnu) (Butler et al. 2011) (PDB ID 3P49). The two molecules of glycine are depicted in brown, and the α and β A-minor interactions are highlighted in green and red. (C) The interface between the aptamers, showing the α, β, γ, and δ interactions, which link the two ligand-binding sites.

The two aptamers in the tandem riboswitch each bind a separate molecule of glycine, but, because of the tandem arrangement, the binding sites are not necessarily independent (Mandal et al. 2004). Tandem riboswitch architectures that act as “genetic logic gates” have been reported (Sudarsan et al. 2006). However, most of these tandem configurations comprise two complete riboswitches, including an expression platform for each aptamer element, so they function independently (Welz and Breaker 2007). For many years, the tandem glycine riboswitch was considered a unique cooperative RNA system (Mandal et al. 2004), behaving as a digital sensor for the concentration of glycine.

The recent identification of a leader helix (boxed in Fig. 1A), the inclusion of which eliminates cooperativity between the aptamer domains, has reopened the debate over the purpose of the tandem structure of the glycine riboswitch (Kladwang et al. 2012; Sherman et al. 2012). A recent calorimetry study, while showing that the leader helix promotes ligand binding and riboswitch compaction, has further questioned whether the tandem riboswitch with the leader P0 helix binds one or two equivalents of glycine (Baird and Ferré-D'Amaré 2013).

The glycine riboswitch lacking the P0 helix has been structurally characterized in both the tandem form (Butler et al. 2011) and as a single aptamer (Huang et al. 2010). The tandem riboswitch forms a semisymmetric dimer, with each aptamer domain binding a separate molecule of glycine in a bulge within helix P3 (Fig. 1B). An extensive network of interaptamer interactions, largely mediated by A-minor contacts between the P1 of one aptamer and the P3 of the other, form an interface between the two ligand-binding sites (Fig. 1C). The single-aptamer construct formed a homodimer in the crystals that closely matched the interaptamer interface in the tandem aptamer structure (Huang et al. 2010). While the discovery of the leader helix has reopened the question of cooperativity, modeling indicates that the kink-turn and P0 helix can be accommodated into the existing structural model (Kladwang et al. 2012). Therefore, the binding-site and interface interactions predicted by the structure remain relevant to understanding this intriguing regulatory system.

Investigations of the glycine riboswitch have relied on assays that indirectly track riboswitch structural changes. However, because the aptamers dimerize, glycine binding in either aptamer is predicted to propagate structural changes throughout both aptamers. We have used an equilibrium dialysis-based assay that directly monitors glycine binding in each aptamer with a set of glycine-binding-site and interface mutants to understand the relationship between dimerization and glycine binding. We propose a revised model for tandem riboswitch function that is consistent with these results, wherein ligand binding in aptamer-1 is linked to aptamer dimerization and stabilizes the P1 stem of aptamer-2, which controls the expression platform.

RESULTS

The tandem glycine riboswitch from Vibrio cholerae, including the leader sequence (VC1-2) was the focus of this study (see Supplemetal Material for full sequence). This particular glycine riboswitch controls the VC1422 gene, which encodes a sodium/alanine- or glycine-symporter. It was the initial glycine riboswitch characterized (Mandal et al. 2004) and has served as the prototype for subsequent biochemical study (Lipfert et al. 2007; Kwon and Strobel 2008; Erion and Strobel 2011). For clarity, the positional nucleotide numbering remains consistent with previous reports, with the leader nucleotides assigned positions -7 to -1 (Sherman et al. 2012; Esquiaqui et al. 2014).

Glycine binding by the wild-type full-length RNA

To investigate ligand binding by the tandem glycine riboswitch independent from any conformational changes, we used a binding assay that directly measures glycine binding using equilibrium dialysis. This analysis was enabled by the inclusion of the leader, which allows the riboswitch to behave well at higher RNA concentrations. We confirmed that wildtype (WT) VC1-2 containing the leader sequence binds glycine noncooperatively with low micromolar affinity, in agreement with other recent reports (Kladwang et al. 2012; Sherman et al. 2012; Baird and Ferré-D'Amaré 2013). Because the inclusion of the leader sequence dramatically changes the ligand-binding activity of the riboswitch, we wished to verify if each ligand-binding site in the tandem riboswitch with leader binds to a separate molecule of glycine.

Glycine binding to in vitro transcribed WT VC1-2 was monitored by equilibrium dialysis (see Materials and Methods), and the fraction of bound glycine increases with RNA concentration. The data were fit well by a standard binding curve (Hill coefficient of 1) with an equilibrium dissociation constant (K_d) of 2.0 µM (Table 1 and Fig. 2C). The measured ligand-binding affinity is in agreement with previous reports, and the Hill coefficient corroborates the recent findings that glycine binding is not cooperative in riboswitches that include the leader sequence. The data fit equally well to a one- or two-site binding model, indicating that either the two sites have very similar affinities or binding to the second site is weak or nonexistent (Fig. 2D). If the two sites are assumed to have equivalent affinities, these data are consistent with two binding events with K_ds of 4 µM.

TABLE 1.

Glycine-binding affinity and equivalents bound of WT VC1-2 and its binding-site mutants

FIGURE 2.

Binding-site mutants of the tandem riboswitch bind a single molecule of glycine in the unmutated aptamer. (A) Detailed structure of the glycine-binding site from the Fnu glycine riboswitch (Butler et al. 2011) (PDB ID 3P49), showing the uracil that contacts the ligand. (B) Secondary structures of the tandem riboswitch showing the U to A mutations that disrupt ligand binding. (C) Glycine binding by WT VC1-2 and its binding-site mutants, showing that the singly binding U78A (Lig2, fuchsia) and U207A (Lig1, blue) bind with near wild-type affinity while the doubly mutated U78A/U207A (brown) shows no binding activity. Here, ligand binding by WT VC1-2 is fit using a single-binding-site model. (D) WT VC1-2 binds to glycine approximately twofold more tightly than predicted for two sites with the Lig1 and Lig2 affinities of 8.5 and 3.7 μM (dashed line). The predicted binding curve for two sites with equivalent affinities of 4.0 μM is shown for comparison (dotted line).

To test whether binding occurs in both sites, we repeated the binding assay with a known excess of glycine. At RNA concentrations ∼30 times the K_d, the WT tandem riboswitch containing the leader sequence binds 1.8 equivalents of glycine (Table 1), in agreement with the original studies on the glycine riboswitch, but in contrast to the recent report of a single equivalent bound as reported using isothermal titration calorimetry (ITC) (Baird and Ferré-D'Amaré 2013). The disparity might arise from a variety of differences between the techniques, including longer equilibration times or refolding the riboswitch in the presence of the ligand. However, we obtained the same results when glycine was added to prefolded WT VC1-2. When binding was monitored at even higher RNA and ligand concentrations (∼100 times the K_d), the tandem riboswitch binds a full two equivalents of glycine. Our results demonstrate that both binding sites are able to bind ligand in the tandem riboswitch with leader.

Mutations to create tandem riboswitches with a single glycine-binding site

Because of the evolutionary conservation of the tandem structure of glycine riboswitches, the presence of two aptamers is expected to contribute to ligand binding or gene control. However, it was not known if glycine binding by both aptamers is necessary to achieve high-affinity binding to either aptamer. Mutation of a conserved guanosine (G) in the three-helix junction of one aptamer has been shown to have no effect on the glycine-binding affinity of the other aptamer, as monitored by in-line probing (Sherman et al. 2012). However, based on the current structural model, these Gs do not directly contact the ligand (Butler et al. 2011), and the effects of these mutations on each aptamer's structure and ligand-binding capacity are unclear. In order to investigate the interdependence of the two aptamers in the tandem structure, we directly mutated the ligand-binding site and determined if disruption of ligand binding in one aptamer affects the affinity of the second.

The crystal structure of the F. nucleatum (Fnu) tandem glycine riboswitch includes a conserved uracil (U) in each ligand-binding pocket that directly contacts glycine (U78 and U207) (Fig. 2A; Butler et al. 2011). We hypothesized that substitution with an adenosine (A) would disrupt this interaction and obstruct the pocket, therefore displacing glycine from the binding site. We incorporated this binding-site mutation into VC1-2 aptamer-1 (U78A), aptamer-2 (U207A), or both aptamers (U78A/U207A) (Fig. 2B). These mutants have not been studied previously because, prior to the structure, ligand binding was thought to occur in the three-helix junction rather than the P3 bulge.

The number of glycine-equivalents bound was determined for the binding-site mutants (Table 1). As predicted, when either ligand-binding site is disrupted independently (U78A or U207A), the tandem riboswitch only binds one glycine equivalent. Furthermore, the disruption of both aptamer binding sites (U78A/U207A) abolished glycine binding by the riboswitch. Therefore, these single-site mutations can be used to selectively eliminate a ligand-binding site. They also corroborate the conclusion that the WT tandem riboswitch binds two equivalents of glycine. To draw attention to the occupied ligand-binding site in these singly glycine-binding mutants, we named them according to the unmutated site. Therefore, U78A is referred to as Lig2, while U207A is Lig1.

Having identified mutants that selectively eliminate glycine binding to either of the tandem aptamers, we determined if disruption of ligand binding by one affects the affinity of the second. The single-glycine-binding mutants of VC1-2 were folded in the presence of trace glycine, and the affinity for the unmutated site was determined by equilibrium dialysis (Table 1 and Fig. 2C). When each ligand-binding site is disrupted independently (Lig1 and Lig2), there is at most a twofold effect on ligand binding to the remaining. This is reflected in the predicted binding curve for two sites with affinities matching those of Lig1 and Lig2 (Fig. 2D, dashed line), which deviates only slightly from the data for WT VC1-2 (squares). When both binding sites are disrupted, the tandem riboswitch no longer binds glycine across the range of concentrations tested. Therefore, the ability to bind glycine in one aptamer has only a small effect on the other aptamer's ligand-binding site.

Because the double-binding-site mutant does not bind glycine at any of the tested RNA concentrations, and because each single mutant binds only one equivalent of glycine, we conclude that the Lig1 and Lig2 constructs only bind ligand in the unmutated ligand-binding site. Therefore, with these constructs it is possible to monitor ligand binding in a specific ligand-binding site, in the context of the full-length tandem riboswitch.

Mutations that disrupt the aptamer–aptamer interface

Dual ligand binding by the tandem riboswitch is not necessary for high-affinity ligand binding and cannot explain the evolutionary conservation of the tandem riboswitch. An alternative explanation for the tandem riboswitch invokes aptamer dimerization as a requirement for ligand binding. Therefore, we tested if the aptamers need to dimerize in order to bind ligand.

The crystal structure of the tandem riboswitch identified a series of tertiary interactions between the two aptamers (Figs. 1C, 3A; Butler et al. 2011). These interactions create an interface between the two ligand-binding sites and include two pseudo-symmetric series of A-minor interactions (the α and β interactions), a cis Hoogsteen base pair (the γ interaction), and a pair of adenosines that flip out of the ligand-binding sites and stack against the three-helix junction and the P1 helix of the cis aptamer (here labeled δ interactions).

FIGURE 3.

Structure-guided perturbation of the aptamer interface. (A) Secondary structure of VC1-2, showing the ligand-binding site and interface residues. Glycine is depicted in brown. The α, β, γ, and δ interactions are based on the crystal structure, shown in Figure 1. Interface positions mutated in this study are shown in bold colors. The L3TL mutation replaces the L3 loop of aptamer-1 with a UUCG tetraloop (black box). (B) Split aptamer constructs used to probe aptamer dimerization in trans by gel-shift electrophoresis. (C) Aptamer–aptamer dimerization curves of interface mutants. WT is shown in black, α mutants in green, β in red, γ in orange, and δ in purple. (D) Sample gel shift showing a tightly associated dimer (A73C) and a weakly associated dimer (U74C).

Mutations were incorporated into the riboswitch to disrupt these interactions (Fig. 3A), as discussed below. The effects of the mutations on dimerization were determined by measuring the affinity between the two aptamers in a trans gel-shift assay (Erion and Strobel 2011; Sherman et al. 2012), with the aptamer-1 portion retaining the 5′-half of the P0 helix and the aptamer-2 construct containing the 3′-half (Table 2 and Fig. 3B,C). Dimer formation was monitored in the presence of saturating glycine by the appearance of a higher molecular-weight complex during gel electrophoresis (Fig. 3D).

TABLE 2.

Aptamer-1/aptamer-2 dimerization affinities in trans of interface mutants of VC1-2

Disruption of the predicted tertiary interactions generally weakens dimerization of the aptamers. The α and β interactions consist of four and three A-minor interactions, respectively, between the P1 stem of one aptamer and a loop or bulge in P3 of the other aptamer. A-minor interactions are disrupted by mutation of the adenosines to cytosines or by the formation of a wobble pair in the P1 stem (Doherty et al. 2001; Kwon and Strobel 2008; Erion and Strobel 2011). We concentrated on the Type I A-minor interactions formed at the top of each P1 stem. At this position in the α interaction, A202 contacts G14:C125 (in the corresponding β interaction, A73 contacts G145:C220). The A202C mutation weakens dimerization of the two aptamers eightfold. The corresponding A73C mutation in the β interaction did not cause similar weakening of dimerization (see Discussion). However, C220U, which forms a wobble pair at the position where A73 contacts the P1 stem, weakens dimerization in trans eightfold. C220U's effects on dimerization cannot be isolated to just the β interaction because G145:C220 is the base pair against which A171 and A219 stack in the δ interaction. However, C220U causes a fourfold larger disruption of dimerization than the A171U mutation, which lends support to the importance of the β interaction for aptamer dimerization. We also created a more extreme mutation L3TL, where L3 of aptamer-1 was mutated to a UUCG tetraloop, which is predicted to disrupt both the β and γ interactions. This mutation dramatically weakens dimerization, causing a 180-fold loss in dimer affinity.

The γ interaction is an A–U Hoogsteen base pair at the center of the dimer interface. Mutation of either the A to a G or the U to a C should disrupt the interaction. While these mutant constructs could rearrange to allow wobble Hoogsteen pairs or wobble Watson–Crick pairs, either possibility would require remodeling of the interface to accommodate the different pairing distances, which would disturb the arrangement of the α and β interactions. U74C causes a 19-fold loss in dimer affinity, while A203G reduces the affinity by 89-fold. These mutations provide the greatest disruption of dimerization of any of the point mutants tested.

The adenines that form each δ interaction are flipped out of the ligand-binding pockets to stack against the three-helix junction and the P1 stem of the cis aptamer. Replacing the adenosines with pyrimidines may weaken the stacking interactions, which will affect P1 stability and could have long-range consequences, given the participation of the P1 stems in the α and β interactions. The A64U and A171U mutations cause 15-fold and twofold weaker aptamer dimerization, respectively. As with the A73C mutation above, the A171U mutation, also located at the top of P1 of aptamer-2, is less detrimental than the equivalent mutation in aptamer-1. While the differences in the two δ mutations’ effects were unexpected, both can be used to perturb dimerization to varying degrees.

The observed effects of the interface mutations on aptamer dimerization in trans confirm the interactions proposed in the structure of the tandem glycine riboswitch. These mutants can be used to perturb interface interactions and disrupt dimerization.

Mutations that disrupt dimerization weaken glycine binding in aptamer-1

We used the interface mutations to determine the importance of aptamer dimerization for glycine binding by the riboswitch. Because the Lig1 and Lig2 constructs cannot bind ligand in the mutated ligand-binding site, these constructs can be used to monitor glycine binding specifically in the remaining functional aptamer in the context of the full-length tandem riboswitch. By combining the binding-site mutations with interface mutations, we tested the importance of aptamer dimerization for the ligand-binding affinity of each binding site.

We determined the glycine-binding affinity of a series of double mutants of VC1-2, each combining an interface mutation with a binding-site mutation (Table 3 and Fig. 4A,B). The affinity of each double-mutant was compared with the affinity of the parent RNA containing only the glycine-binding-site mutation. In each case tested, the interface point mutations to α, β, and γ interactions disrupt the glycine-binding activity of the Lig1 constructs, but have little to no effect on binding by the Lig2 constructs (Fig. 4B). Since the Lig1 constructs only harbor an active aptamer-1 ligand-binding site, these results indicate that disruption of the dimer interface disproportionately impacts ligand binding in aptamer-1.

TABLE 3.

Glycine-binding affinities of interface/binding-site double mutants

FIGURE 4.

Mutations that disrupt the dimer interface have little effect on glycine binding by aptamer-2, but disrupt glycine binding by aptamer-1 to a degree proportional to their effect on dimerization. (A) Sample binding curves for U74C mutants. Parent binding-site mutants are shown in black, U74C mutants in orange. (B) Fold change in glycine-binding affinity of an interface mutant relative to the single-glycine-binding parent, for Lig1 (dark gray) and Lig2 (light gray). (C) Comparison of various ligand-binding models to the U74C glycine-binding data. The one-site model (dotted line, 9.2 μM K_d) and two-site model (solid line, 9.2 and 300 μM K_ds) are largely superimposed and fit the data well (R² > 0.995). A predicted two-site binding curve based on the affinities measured in the Lig1 and Lig2 backgrounds (dashed line, 230 and 4.8 μM K_ds) would have approximately threefold tighter affinity than the measured binding data. (D) Comparison of interface mutations’ effects on dimerization and glycine binding by aptamer-1. Mutations that disrupt the α interaction are shown in green, β in red, γ in orange, and δ in purple. The L3TL mutation disrupts both the β and γ interactions. The dashed line illustrates a trend and is not a fit. (#) Lig1 L3TL shows no detectable binding at 500 μM glycine, which is at least 75-fold weaker than the parent construct.

For example, when either side of the γ interaction is mutated (U74C or A203G), ligand binding is only slightly affected in the Lig2 constructs (about twofold). In contrast, the identical γ mutations reduce ligand binding by 27- and 72-fold, respectively, for the Lig1 constructs. Similarly, mutation of the α (A202C) or β (A73C or C220U) interaction has little or no effect on ligand binding in aptamer-2 but significantly reduces ligand binding in aptamer-1 (10-, 37-, and 36-fold, respectively). The L3 tetraloop mutation (L3TL), designed to disrupt both the β and γ interactions, has the greatest effect on glycine binding by aptamer-1 (undetectable binding, which is a >75-fold change). This extreme mutation weakens glycine binding by aptamer-2 by a more modest 11-fold.

Disruption of the δ interactions does not fully follow the pattern identified above, perhaps because the adenosine involved in the δ interaction is located in the glycine-binding pocket. Mutation of the δ adenosine in aptamer-1 (A64U) has a significant effect (62-fold) on glycine binding in aptamer-1 with little to no effect on glycine binding in aptamer-2. Mutation of the δ adenosine in aptamer-2 (A171U) has a moderate effect on glycine binding in both aptamers (nine- and fourfold).

We plotted the effect of each interface mutation on aptamer dimerization in trans against its effect on glycine binding in aptamer-1 (Fig. 4D). Not only do mutations that disrupt dimerization show significantly weaker ligand binding in aptamer-1, these effects are directly proportional, indicating a linkage between the two equilibria. The mutants that fall below the line are A64U and the β mutants. A64 is located in the ligand-binding site of aptamer-1, and so it is reasonable that A64U would affect ligand binding by aptamer-1 in excess of the amount predicted by its effect on dimerization. Indeed, the equivalent mutation in aptamer-2, A171U, disrupts ligand binding by aptamer-2 ninefold, the only point mutant to have such an effect. Therefore, once the effects on the cis binding site are considered, the A64U mutation exhibits proportional effects on glycine affinity and dimerization. It should be noted that these two affinities, ligand binding and dimerization, are not independent (Sherman et al. 2012 and results below). This analysis illustrates a trend, rather than quantitatively characterizing a dependence. In general, disruption of the dimer interface weakens ligand-binding affinity, particularly for aptamer-1, which is weakened proportionally.

Ligand-binding affinities in WT-binding-site background

We next determined if the effects of interface mutations on ligand binding in the Lig1 and Lig2 backgrounds were indicative of the effects of those same mutations on ligand binding in the wild-type-binding-site background. We determined the glycine-binding activity of the interface mutants in RNAs with WT glycine-binding sites across a range of RNA concentrations (Table 4). As with the fully WT VC-12, data for interface mutants in the WT-binding-site background fit equally well to a one- or two-site model (Fig. 4C). In some cases, the affinity measured in the WT background was as strong as the tightest single-binding site measurement (A73C, U74C, A64U, A171U, L3TL) and was presumably dominated by binding in aptamer-2. In other cases, the affinity measured in the WT background approached the average of the two affinities measured in the single-binding site backgrounds (A202C, C220U, A203G).

TABLE 4.

Glycine-binding affinities of interface mutants in WT-binding-site background

Given the difficulties of deconvoluting the two binding sites in experiments performed with trace glycine, we next analyzed the interface mutants in the WT-binding-site background using equilibrium dialysis experiments with excess glycine. Under these conditions, ligand binding in the weaker binding site can be observed. Using the experimentally determined K_d values for the two ligand-binding sites in the Lig1 and Lig2 backgrounds, we predicted the binding site occupancy at RNA and glycine concentrations in the dynamic region of the binding curve (Supplemental Material). We then experimentally determined the equivalents of glycine that were bound by the tandem riboswitch at that same concentration (Table 5).

TABLE 5.

Glycine-equivalents bound by VC1-2 interface mutants at the indicated RNA and glycine concentrations

For example, the mutant Lig2 U74C has a glycine affinity of 4.8 μM, and consequently, aptamer-2 is predicted to be >99% occupied at 140 μM RNA and 380 μM glycine. In contrast, Lig1 U74C has an affinity of 230 μM; thus, aptamer-1 is predicted to be 20% occupied under those same conditions. When these values are adjusted to reflect the approximately twofold tighter ligand binding in the wild-type-binding-site background (see Supplemental Material for full calculations), aptamer-1 is predicted to be 40% occupied when the U74C mutation is introduced into a WT-binding-site background. VC1-2 μM binds 1.2 equivalents of glycine at these concentrations of RNA and ligand, which is consistent with the expected range of 1.2–1.4 equivalents.

As shown in Table 4, glycine binding in the WT-binding-site background is consistent with the affinities measured in the Lig1 and Lig2 backgrounds for most of the interface mutants. The outliers are A73C and C220U, which disrupt the β interaction, and A203G, which disrupts the γ, all of which bind ∼20% more glycine in the WT background than we would predict based on the affinities measured in the single-site backgrounds (see Discussion). In general, the behavior of interface mutations in the Lig1 and Lig2 backgrounds is a good predictor for their behavior in the WT-binding-site background. In the tandem glycine riboswitch, disruption of the dimer interface weakens ligand-binding affinity, particularly for aptamer-1.

Aptamer-2 constructs only bind glycine when able to form a homodimer

As described above, interface mutations have little to no effect on ligand binding by aptamer-2 in the context of the tandem riboswitch. One potential explanation for this observation is that aptamer-2 is functional as a monomer. Consistent with this hypothesis, an aptamer-2 construct, VC2 (Fig. 5A), has been shown to bind glycine with reasonably high affinity (Mandal et al. 2004; Erion and Strobel 2011). However, structural (Huang et al. 2010) and biochemical analysis of this and similar constructs indicate that they form a homodimer with a K_d value for dimerization close to the concentration at which those ligand-binding experiments were performed. Another aptamer-2-only construct, VC2s (Fig. 5A), which lacks the linker region, is active when aptamer-1 is added in trans but has never shown glycine-binding activity in isolation (TV Erion, pers. comm.). A third aptamer-2-only construct, which included the linker region, showed barely detectable glycine-binding affinity in isolation, but addition of aptamer-1 in trans rescued near WT affinity (Sherman et al. 2012).

FIGURE 5.

Aptamer-2 constructs that bind glycine function as homodimers, and mutations that disrupt homodimerization also disrupt glycine binding (green). (A) Aptamer-2 constructs tested for homodimerization and glycine binding. VC2 homodimerization is likely mediated by a helix formed between two VC2 “tails.” (B) Homodimerization affinity of aptamer-2 constructs. (C) Glycine-binding affinity of aptamer-2 constructs. (D) VC2 homodimerizes in a glycine-independent manner, while VC2s requires glycine for homodimerization, as seen by gel shift in saturating glycine (closed symbols) or an equivalent concentration of alanine (open symbols).

In order to determine if homodimerization is important for glycine binding by the aptamer-2-only constructs, VC2 and VC2s, we mutated all three adenosines involved in type-I A-minor interactions in the α interaction and then determined if this affected the homodimerization and glycine-binding affinities (Table 6 and Fig. 5). In order to monitor homodimerization, a small amount of radiolabeled aptamer-2 construct was mixed with an excess of the same RNA species and refolded in the presence of saturating glycine. Dimer formation was observed by the appearance of a higher molecular-weight complex during gel electrophoresis. The glycine-binding affinity of the aptamer-2 constructs was determined by equilibrium dialysis of a mixture of RNA and trace, radiolabeled glycine.

TABLE 6.

Homodimerization and glycine-binding affinities of aptamer-2-only constructs

The aptamer-2 construct that includes the linker region and the last several nucleotides of aptamer-1, VC2, binds glycine with 35 µM affinity (Fig. 5C), consistent with previous reports. This construct forms a homodimer with 200 nM affinity in saturating glycine. Attempts to render VC2 monomeric were unsuccessful (Fig. 5B), and VC2 homodimerizes at approximately the same affinity in saturating glycine and in alanine (Fig. 5D). This ligand-independent dimerization is likely mediated by a 10 bp helix, which can form between the “tails” of two VC2 constructs (Fig. 5A). This 10 bp helix is consistent with the recent report of construct inhibition when aptamer-2-only constructs are lengthened to include pairing regions that would interfere with its formation (Sherman et al. 2014). Therefore, VC2 forms an obligate homodimer and cannot be used to determine if monomeric aptamer-2 can bind glycine.

VC2s, the aptamer-2 construct lacking the linker region, binds glycine much more weakly than VC2 (Fig. 5C). The glycine-binding curve for VC2s is best fit by a model for cooperative ligand binding with a Hill coefficient of 1.8 and an equilibrium constant equal to 300 μM, consistent with the ligand-binding equilibrium depending on two molecules of aptamer-2, likely because of homodimerization. VC2s WT forms a homodimer in saturating glycine with an affinity of 10 μM (Fig. 5B), and VC2s dimerization is ligand dependent (Fig. 5D). The glycine-binding curve for VC2s probably corresponds to homodimerization, even though the affinity is >35-fold weaker than that measured by gel shift, because the gel shift occurs in saturating glycine while ligand binding was studied in trace glycine. Mutating all three adenosines involved in type-I A-minor interactions in the α interaction disrupts dimerization entirely at the concentrations tested, a >45-fold loss in dimer affinity (Fig. 5B). This VC2s α mutant shows barely detectable glycine binding at the highest RNA concentration tested (Fig. 5C). Therefore, aptamer-2 is not able to function in isolation as a monomer.

Aptamer dimerization depends on ligand binding

The decrease in ligand-binding affinity of VC1-2 and VC2 upon disruption of the dimer interface indicates that dimerization is important for glycine binding. Therefore, the reciprocal dependence should also be present, where ligand binding promotes dimerization of the two aptamer domains into a well-folded tertiary structure. To test this interdependence, we analyzed the dimerization affinity of the two aptamer domains in trans in the absence of glycine binding, either by disrupting the ligand-binding sites or performing gel shifts in the presence of alanine in place of glycine (Table 7 and Fig. 6). In both cases, the dimer affinity weakens by >80-fold, indicating that the tandem riboswitch requires glycine binding to form a well-folded dimer. Interestingly, the dimerization affinity of the L3TL mutant in both glycine and alanine is similar to that of the WT construct in alanine, indicating that the residual affinity is not dependent on the aptamer–aptamer interface. We attribute the residual ∼20 μM affinity to base-pairing within the P0 helix, which is split between the two constructs.

TABLE 7.

Aptamer-1/aptamer-2 dimerization affinities in trans when glycine binding is disrupted

FIGURE 6.

Aptamer dimerization depends on glycine binding. (A) When both ligand-binding sites are disrupted (U78A/U207A, in brown) or no ligand is present (alanine, open symbols), dimerization is substantially weakened. The L3TL interface mutation (red) does not further weaken the affinity. (B) Disrupting the ligand-binding sites singly (Lig1, blue and Lig2, fuschia) has an intermediate effect on dimerization.

When a single ligand-binding site is disrupted, dimer formation in trans is weakened but not abrogated. The Lig1 and Lig2 constructs dimerize in saturating glycine with nine- and fivefold weaker affinities than WT, but still >10-fold stronger than the no-glycine cases. Therefore, ligand binding in each binding site contributes to promoting dimerization of the two aptamer domains into a well-folded tertiary structure, and ligand binding and aptamer dimerization are linked equilibria.

DISCUSSION

In this study, we examined two aspects of the tandem riboswitch that might provide selective advantage over a single-aptamer system: double binding-site occupancy and aptamer dimerization. We show that double binding-site occupancy is not necessary for high-affinity ligand binding. In contrast, aptamer dimerization is energetically linked to ligand binding, particularly in aptamer 1. Based on our results, we propose a model for riboswitch function (Fig. 7), wherein ligand binding in aptamer-1 is linked to aptamer dimerization and stabilizes the P1 stem of aptamer-2, which controls the expression platform.

FIGURE 7.

Proposed model for glycine binding by the tandem glycine riboswitch. In the absence of glycine (top), dimerization of the two domains is disfavored. In this case, the P1 stem of aptamer-2 is largely not formed, instead interacting with the downstream expression platform (red domains form an alternative helix). Upon addition of glycine (bottom), the equilibrium is shifted toward aptamer dimerization, with the P1 stem of aptamer-1 providing a scaffold for the dimer interface. Dimerization stabilizes the P1 stem of aptamer-2, which modulates the structure of the downstream expression platform.

Our analysis confirms that the tandem glycine riboswitch from V. cholerae containing the leader sequence noncooperatively binds glycine with low micromolar affinity, in agreement with other recent reports (Kladwang et al. 2012; Sherman et al. 2012; Baird and Ferré-D'Amaré 2013). This lack of cooperativity has reopened the debate over the purpose of the tandem structure of the glycine riboswitch. Because the tandem architecture has been conserved against evolution, rather than being reduced to a simpler single-aptamer riboswitch, it is expected to provide some benefit in ligand-binding affinity, kinetic response time, or complex genetic control.

In contrast to a recent binding study using isothermal titration calorimetry (Baird and Ferré-D'Amaré 2013), we show that the tandem glycine riboswitch binds two equivalents of glycine. The disparity does not appear to result from technical differences between the techniques. Given the relationship that we have demonstrated between ligand binding and dimerization (Fig. 6B), we speculate that the heat evolved upon ligand binding in the ITC experiments results from dimerization, which is significantly promoted by binding of the first equivalent of ligand. It is worth noting that many transcriptionally controlled riboswitches are not under thermodynamic control (Serganov and Patel 2012) and references therein), so neither technique entirely describes the riboswitch–ligand interaction. However, our results demonstrate that both binding sites are able to bind ligand in the tandem riboswitch with leader.

Because Lig1 and Lig2 bind almost as well as WT, the two ligand-binding sites are independent. This analysis assumes that the U-to-A binding-site mutation displaces the glycine without stabilizing the binding site in a “bound-like” conformation. As a counter-example, the C64U mutant of the adenine riboswitch fails to respond to ligand, instead causing constitutive activation of the downstream gene (Tremblay et al. 2011). Given the significant disruption of dimerization when both binding sites are mutated, and the similarity of U78A/U207A's dimer affinity to that of the WT constructs in the absence of ligand (Fig. 6), the U-to-A mutations behave like empty sites. Therefore, dual ligand binding by the tandem riboswitch is not necessary for high-affinity binding and cannot explain the evolutionary conservation of the tandem riboswitch.

An alternative explanation for the tandem riboswitch invokes aptamer dimerization as a requirement for ligand binding. Mutations to the predicted interface interactions disrupt aptamer dimerization in trans in a manner largely consistent with the structural model (Fig. 3). The exceptions cluster around the top of the P1 helix in aptamer-2. Given the considerable consequences of the β mutants for glycine binding by aptamer-1, we propose that remodeling of the interaction conceals the effects of these mutations on aptamer dimerization, as discussed further below. We also studied a more extreme mutation, L3TL, which should disrupt both the β and γ interactions. Because this mutant's dimerization affinity was equally poor in the presence and absence of glycine (Fig. 6), we consider L3TL to completely disrupt the interface. L3TL and the point mutants provide a range of interface mutations that can be used to disrupt aptamer dimerization.

We used these interface mutations to determine the importance of aptamer dimerization for glycine binding by the riboswitch. Intriguingly, glycine binding by aptamer-1 is much more sensitive to dimerization than glycine binding by aptamer-2. In each case tested, the interface point mutations to α, β, and γ interactions all disrupted the glycine-binding activity of the Lig1 constructs, while having little to no effect on binding by the Lig2 constructs (Fig. 4C). The L3TL mutant, which disrupts both the β and γ interactions, has the largest effect on ligand binding in aptamer-1. In addition, this extreme interface mutation also weakens ligand binding in aptamer-2. Therefore, both aptamers require a well-folded dimer interface for ligand binding, but aptamer-1 is much more sensitive to perturbations of that interface.

Furthermore, each interface mutation's effect on ligand binding in aptamer-1 and dimerization are directly proportional (Fig. 4D), indicating a linkage between the two equilibria. The mutants that fall below the line are A64U and the β mutants. A64 is located in the ligand-binding site of aptamer-1, and once the effects on the cis binding site are considered, the A64U mutation exhibits proportional effects on glycine affinity and dimerization. The β mutants, A73C and C220U, both significantly disrupt ligand binding in aptamer-1 while having little or no effect on aptamer dimerization affinity in trans. In addition, A73C and C220U are two of the three mutants that bind glycine better in the WT-binding-site background than would be predicted based on the measured values in the Lig1 and Lig2 backgrounds. These discrepancies suggest there is remodeling around the top of the P1 stem of aptamer-2 when mutations disrupt the β interaction. As this stem is predicted to change conformations between the on- and off-states (Mandal et al. 2004), interactions that stabilize the appropriate conformation could have important consequences for gene control beyond their effects on ligand-binding affinity.

At least two possible models of riboswitch function could account for the asymmetric effect that disrupting dimerization has on ligand binding by the two aptamers. The simplest explanation would be that aptamer-2 is able to function as a monomer, while aptamer-1 requires dimerization for ligand binding to occur. We tested the homodimerization and glycine-binding affinities of two aptamer-2-only constructs (Fig. 5), and showed that aptamer-2 is not able to function in isolation as a monomer.

Our proposed model for riboswitch function, outlined in Figure 7, invokes aptamer dimerization as a key modulator of P1 formation in aptamer-2. Therefore, the tandem glycine riboswitch can be considered an extreme example of “inverse junctional architecture” (Serganov and Patel 2012; Serganov and Nudler 2013), wherein ligand binding affects P1 stability through stabilization of global conformation, including the formation of long-range tertiary interactions between the two domains. This model presumes that dimerization and ligand binding are linked equilibria and predicts that dimerization of the two aptamers in trans is weakened when ligand binding is impaired. We analyzed the dimerization affinity of the two aptamer domains in trans in the absence of glycine binding (Fig. 6A), and the dimer affinity weakens significantly, indicating that the tandem riboswitch requires glycine binding to form a well-folded dimer. When a single ligand-binding site is disrupted, dimer formation in trans is weakened but not abrogated (Fig. 6B). Therefore, ligand binding in each binding site contributes to promoting dimerization of the two aptamer domains into a well-folded tertiary structure, with the interdependence between ligand binding and dimerization particularly strong for aptamer-1.

This model proposes that, in order for ligand binding to occur in either aptamer, both aptamers must form the dimer interface. The asymmetric effects on ligand binding by the two aptamers could be explained by differences in the relative stabilities of the P1 stems, as the P1/P0 helix of aptamer-1 is twice as long as that of aptamer-2. The predicted folding energies for helices with these lengths and sequences show the P1/P0 helix of aptamer-1 is 8 kcal/mol more stable than the P1 of aptamer-2 (Zuker 2003). This difference in stem-length is conserved across tandem glycine riboswitches, with aptamer-1 always containing a 3–7 base P0 with an 8–9 base P1, while aptamer-2 has only a 5–6 base P1 helix (Mandal et al. 2004; Sherman et al. 2012). Similar differences in P1 helix length and stability have been shown to cause pronounced differences in structure and ligand responsiveness for two adenine riboswitches (Nozinovic et al. 2014).

We propose that this additional stability causes the P1 of aptamer-1 to form independently of ligand binding. In-line probing experiments have shown that the P1 and P0 helices of aptamer-1 are protected in the presence and absence of glycine (Sherman et al. 2012), and in recent spin labeling experiments, P0 is largely formed upon addition of monovalent cations (Esquiaqui et al. 2014). In contrast, aptamer-2's P1 is predicted to undergo a conformational change on ligand binding, favoring aptamer formation over an alternative helix with the expression platform, although, in experiments performed without the competing expression platform, aptamer-2's P1 does not show differential reactivity in the presence and absence of ligand (Mandal et al. 2004).

Because aptamer-1's P1 stem is preformed, aptamer-2 is able to take advantage of aptamer-1's structure and more readily form dimer, even when the tertiary interface is partially disrupted. In contrast, aptamer-1 requires an intact tertiary interface in order to constrain aptamer-2 in a bound-like dimeric structure, particularly when ligand binding in aptamer-2 is disrupted. Functionally, this asymmetry in P1 stability could allow aptamer-1 to scaffold the dimer interface, which forms along the P1 stems of the two aptamers.

Other riboswitches have been shown to use scaffolding to preform significant portions of the aptameric secondary structure prior to ligand binding, including prequeuosine class-II (Soulière et al. 2013), S-adenosylmethionine (SAM)-I (Heppell et al. 2011), SAM-II (Haller et al. 2011), cyclic-di-guanosine monophosphate (c-di-GMP) (Wood et al. 2012), and the purine riboswitches (Lemay et al. 2006; Brenner et al. 2010; Nozinovic et al. 2014). Scaffolding can have important consequences for kinetically controlled riboswitches (Wickiser et al. 2005a,b; Trausch and Batey 2014), allowing ligand binding to occur on transcriptionally relevant time scales. As some glycine riboswitches control transcription termination, they are likely to be kinetically controlled (Serganov and Patel 2012).

In contrast, many glycine riboswitches in γ proteobacteria control translation (Mandal et al. 2004), and these riboswitches are likely to be thermodynamically controlled (Rieder et al. 2007; Lemay et al. 2011; Serganov and Patel 2012). In these cases, ligand binding in aptamer-1 could promote dimerization. Because many of the tertiary interactions involve the P1 stem of aptamer-2, dimerization could stabilize the P1 switch, providing extra energy not provided by the binding of the small ligand (Zhang et al. 2014). In addition, based on many other riboswitch systems, it is likely that ligand binding in aptamer-2 directly stabilizes the cis P1 stem. In this way, binding events in each aptamer could independently stabilize the P1 switch, with dimerization and the interface relaying energy from the binding site in aptamer-1.

While the tandem glycine riboswitch remains the only riboswitch system with two homologous aptamer domains that regulate a single expression platform, another riboswitch has recently been shown to bind two ligand molecules within a single-aptamer domain, tetrahydrofolate (THF) (Trausch et al. 2011; Trausch and Batey 2014). In the case of the THF riboswitch, both kinetic scaffolding and thermodynamic cooperativity have been suggested as rationale for dual ligand binding. Several natural variants of the THF riboswitch were analyzed using ITC and structure probing, and, while the Hill coefficients varied, dual ligand binding was conserved. Mutants that disrupt each site were identified and analyzed in one of the cooperative parent constructs. In transcriptional termination assays, the two singly binding mutants of the THF riboswitch diverge significantly, with one continuing to control gene expression, albeit with a moderately reduced effective concentration, while the other mutant fails to respond to ligand. The authors propose that binding at the pseudoknot site is critical for switching, while binding at the distal site could play a scaffolding role and/or allow a cooperative response to changes in THF concentration.

In this study, we demonstrate that aptamer dimerization is energetically linked to ligand binding in aptamer-1. Based on our results, we propose a model for riboswitch function (Fig. 7), wherein (1) both aptamers must adopt a dimeric tertiary structure for ligand binding to occur, (2) aptamer-1 has a stabilized P1 that acts as a scaffold for dimerization, and (3) dimerization and ligand binding stabilize the P1 helix of aptamer-2, which serves as the switch for gene control. Such a model for riboswitch function could explain the prevalence of the tandem riboswitch in two different ways. Dimerization of the two domains could act thermodynamically, providing extra energy to counter-balance the alternative conformation of the expression platform. In contrast or in addition, dimerization could be important kinetically, with scaffolding by aptamer-1 playing an important role in the speed at which aptamer-2 folds and binds ligand.

MATERIALS AND METHODS

DNA oligonucleotides and chemicals

DNA oligonucleotides were synthesized by the W.M. Keck Foundation Biotechnology Resource Laboratory at Yale University and used without further purification. Glycine and other chemicals were obtained from Sigma.

DNA constructs

The V. cholerae VC1422 glycine riboswitch (VC1-2 WT) and single-aptamer constructs were made by adding the seven base leader sequence to previously reported plasmids using a PCR reaction with corresponding primers (Erion and Strobel 2011). The plasmids consisted of the T7 promoter sequence, riboswitch DNA sequence, and the anti-genomic HDV ribozyme sequence in the pUC19 (NEB) plasmid. Mutant riboswitch constructs were made by PCR reaction using corresponding primers.

In vitro transcription

Plasmid DNA encoding the glycine riboswitch was linearized by restriction digest and used as template for transcription by T7 RNA polymerase. RNAs were transcribed in 40 mM Tris–HCL (pH 7.5), 4 mM spermidine, 10 mM DTT, 55 mM MgCl₂, 0.05% Triton X-100, and 4 mM of each 5′-nucleotide triphosphate (7 mM for GTP) for 2 h at 37°C. The HDV ribozyme was allowed to self-cleave by heat denaturing at 70°C and slow refolding in an additional 100 mM NaCl and 20 mM MgCl₂. All RNAs were purified by 6% PAGE, eluted into 0.3 M NaOAc (pH 5.2), precipitated with ethanol, and resuspended in the appropriate buffer. RNA transcripts were then buffer-exchanged four times and concentrated using Amicon Ultra centrifugal filters. RNA concentrations were determined by UV absorbance at 260 nm. Absorption coefficients were determined by digestion with Nuclease P1, according to established protocols (Cavaluzzi and Borer 2004; Wilson et al. 2014). Briefly, ∼1 nmol of RNA was incubated at 50°C for 1 h with 1 unit of Nuclease P1 in 200 mM NaOAc, pH 5.3 with 5 mM EDTA and 10 mM Zn(OAc)₂. Based on extinction coefficients for the individual nucleotides, the extinction coefficient of fully digested VC1-2 WT is 2.7 M⁻¹ cm⁻¹ and that of the intact, folded RNA is 2.0 M⁻¹ cm⁻¹. The extinction coefficients for fully digested and intact, folded aptamer-1 are 1.6 and 1.1 M⁻¹ cm⁻¹, respectively.

Equilibrium dialysis assay

VC1-2 RNA transcripts were combined with trace ¹⁴C-labeled glycine in TB buffer (90 mM Tris-borate at pH 8.3) containing 10 mM MgCl₂ and 100 mM KCl. Samples were heated to 60°C then allowed to slow cool to ∼30°C over an hour. The RNA/glycine mixture was equilibrated overnight at 23°C across from an equal volume of buffer in a 5000 MW cut-off Dispo Equilibrium Dialyzer from Harvard Apparatus. For the highest RNA concentrations tested, osmosis resulted in increased volume on the RNA side of the dialyzer, and the estimated RNA concentration was adjusted accordingly. The amount of ¹⁴C-labeled glycine on each side of the dialyzer was determined by scintillation counting in Ultima Gold on a PerkinElmer Tri-Carb 2910TR scintillation counter. The fraction bound was determined for each sample ([counts on RNA side − counts on buffer side]/counts on RNA side). The K_d value for glycine binding was determined by plotting the fraction bound value versus the concentration of RNA and fitting to a standard equation for one-site binding, using Prism to perform a least squares regression:

where Y is the fraction bound, X is the concentration of RNA, and NS is a constant term for nonspecific binding.

For mutants that failed to saturate at the RNA concentrations tested, B_max values were fixed at 0.98, as indicated in the table legends.

For doubly glycine-binding mutants, the data were also fit to an equation for binding to a two-site model:

where Y is the fraction bound, X is the concentration of RNA, and NS is a constant term for nonspecific binding. In all cases, the data fit equally well to a one-site or two-site model, and the error on fitting the weaker site was >10¹⁰.

For the aptamer-2-only constructs, for which homodimerization was in question, the data were also fit to an equation for cooperative binding:

where Y is the fraction bound, X is the concentration of RNA, and n is the Hill coefficient, which describes cooperativity.

Equilibrium dialysis equivalents assay

VC1-2 RNA transcripts were combined with a threefold excess of cold glycine as well as trace ¹⁴C-labeled glycine. The RNA was refolded and equilibrated with buffer in a 5000 MW cut-off equilibrium dialysis cassette, as discussed above. The amount of ¹⁴C-labeled glycine on each side of the dialyzer was determined by scintillation counting, and the equivalents of bound glycine were determined ([counts on RNA side − counts on buffer side] × 3/[counts on RNA side + counts on buffer side]).

Native gel analysis of aptamer interaction

RNA transcripts were dephosphorylated and 5′-³²P labeled as previously described (Ryder and Strobel 1999). 5′-end labeled aptamer-2 RNA was combined with aptamer-1 RNA constructs at increasing concentrations in TB buffer containing 5 mM glycine, 10 mM MgCl₂, 10 mM KCl, and 10% glycerol. Samples were heated to 90°C for 2 min, then allowed to slow cool from 60°C to ∼30°C over an hour, incubated for an additional 20 min at 23°C, then cooled on ice for 10 min. The reaction mixture was loaded onto a 6% native acrylamide gel in TB buffer containing 10 mM MgCl₂ and 5 mM glycine. Electrophoresis was carried out at 4°C for 2 h in TB buffer containing 10 mM MgCl₂. The separated RNA was visualized by PhosphorImager and bands quantified using ImageQuaNT. For weaker interactions, where the complex's lifetime was <2 h, all signal that shifted above the starting band was considered bound (Fig. 3D). The K_d value for the dimer interaction was determined by plotting the fraction bound value versus the concentration of the aptamer-1 construct and fitting to a standard equation for one-site binding using Prism to perform a least squares regression:

where Y is the fraction bound, X is the concentration of RNA, and NS is a constant term for nonspecific binding. Because many mutants failed to saturate at the RNA concentrations tested, B_max values were fixed at 0.92, the value at which WT saturated.

For the aptamer-2-only constructs, for which homodimerization was in question, the data were fit to a standard binding curve, as our analysis monitored a trace amount of labeled material, which behaves in a pseudo-first order fashion.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.