A Kissing Loop Is Important for btuB Riboswitch Ligand Sensing and Regulatory Control*

Background: The btuB riboswitch contains an unusual kissing loop structure that exhibits few interactions with bound adenosylcobalamin. Results: Structural probing and in vivo analysis show that kissing loop mutations strongly uncouple ligand binding from riboswitch regulation. Conclusion: The kissing loop is pivotal for coordinating riboswitch conformational changes. Significance: In contrast to most riboswitches, btuB relies on a kissing loop to achieve both metabolite sensing and gene regulation. RNA-based genetic regulation is exemplified by metabolite-binding riboswitches that modulate gene expression through conformational changes. Crystal structures show that the Escherichia coli btuB riboswitch contains a kissing loop interaction that is in close proximity to the bound ligand. To analyze the role of the kissing loop interaction in the riboswitch regulatory mechanism, we used RNase H cleavage assays to probe the structure of nascent riboswitch transcripts produced by the E. coli RNA polymerase. By monitoring the folding of the aptamer, kissing loop, and riboswitch expression platform, we established the conformation of each structural component in the absence or presence of bound adenosylcobalamin. We found that the kissing loop interaction is not essential for ligand binding. However, we showed that kissing loop formation improves ligand binding efficiency and is required to couple ligand binding to the riboswitch conformational changes involved in regulating gene expression. These results support a mechanism by which the btuB riboswitch modulates the formation of a tertiary structure to perform metabolite sensing and regulate gene expression.

Riboswitches are highly structured regulatory elements located in untranslated regions of bacterial mRNAs and are involved in controlling gene expression by modulating RNA structure upon metabolite binding (1). These RNA switches exert regulatory control at different genetic levels such as transcription, translation, and mRNA decay (1)(2)(3). Riboswitches are composed of two domains consisting of an aptamer, which is implicated in metabolite binding, and an expression platform, which is involved in gene expression regulation. The vast majority of riboswitches negatively regulate the expression of downstream gene(s) upon metabolite binding (1). Although riboswitch aptamer domains are highly conserved to allow specific interaction with cellular metabolites, expression platforms exhibit both sequence and structural diversity to coordinate the aptamer conformation and genetic modulation. In the case of these transcriptional switches, ligand binding promotes the formation of an anti-antiterminator, the P1 stem, that allows the folding of a terminator structure involved in transcription termination. However, for translationally controlling riboswitches, P1 stem formation promotes sequestration of the ribosome binding site (RBS) 2 sequence, leading to inhibition of translation initiation. Thus, for both transcription-and translation-controlling riboswitches, modulation of the P1 stem secondary structure is central to the outcome of genetic expression (3,4).
Some riboswitches have been reported to rely on the modulation of tertiary interactions such as pseudoknot structures to control gene expression (5). For example, the S-adenosylmethionine II riboswitch modulates the formation of a pseudoknot interaction involved in the sequestration of the RBS sequence when bound to S-adenosylmethionine (6,7). Another example is the adenosylcobalamin (AdoCbl)-sensing riboswitch, which represents one of the most intricate and complex riboswitch structures including a kissing loop (KL) interaction ( Fig. 1A) (8,9). This riboswitch is involved in the control of the btuB gene, which encodes an outer membrane receptor for transporting B 12 derivatives, as part of the TonB-dependent active transport system (10 -12). AdoCbl is a coenzyme that originates from vitamin B 12 and is used in radical chemistry to cleave various covalent bonds (13). Other B 12 derivatives most commonly include methylcobalamin, cyanocobalamin, and aquocobalamin, which differ in their composition by their corrin ring (13). It was recently shown that methylcobalamin and aquocobalamin are also recognized by different cobalamin riboswitches where various RNA architectures are used to sense these B 12 derivatives (9). As reported for the S-adenosyl-* This work was supported in part by the Canadian Institutes of Health Research. The authors declare that they have no conflicts of interest with the contents of this article. 1  methionine II riboswitch, AdoCbl binding to btuB inhibits translation initiation by precluding ribosome access to the RBS sequence (14). A recent phylogenetic analysis is consistent with a switching mechanism for the Escherichia coli btuB riboswitch involving anti-RBS (A-RBS) and anti-anti-RBS (AЈ-RBS) sequences that selectively interact to allow ribosome binding in the absence of AdoCbl (Fig. 1, A and B). The E. coli btuB aptamer domain is organized around a four-way junction exhibiting highly conserved residues (B 12 element) that coordinates a complex secondary structure composed of 12 helical domains (15,16). This junction also encompasses the AЈ-RBS sequence as part of the P1 stem (Fig. 1A). The aptamer domain is also involved in the formation of a kissing loop structure through the interaction of L5 and L13 stem-loop structures (Fig. 1C). Mutagenesis experiments showed that residues forming the kissing loop interaction are important for ligand binding (10), consistent with the kissing loop being involved in genetic control. Recent crystal structures revealed that the bound ligand interacts weakly with the kissing loop ribose-phosphate backbone through van der Waals contacts (8,9). The kissing loop interaction was also shown to form in the absence of ligand at elevated Mg 2ϩ ion concentrations (9). However, AdoCbl binding to the riboswitch was found to reduce the required Mg 2ϩ concentration to physiological levels (9,17). Interestingly, an additional interaction involves the AdoCbl adenosyl moiety and a highly conserved adenine residue located in the AЈ-RBS region of the four-way junction (8,9), providing a mechanism for distinguishing between AdoCbl and related cellular metabolites such as methylcobalamin and aquocobalamin. Although peripheral extensions (P6 -P10 helices) are involved in ligand recognition (8,9,18), there is relatively little information about the role of the kissing loop structure in the folding and ligand-induced genetic regulation of the btuB riboswitch. A recent study reported that transcriptional pause sites are present upstream of the kissing loop (C184), RBS (U225), and AUG sequences (U235) (19). Given that nascent mRNA exiting from paused RNA polymerases coincides with sequences allowing aptamer formation (C184) and kissing loop interaction (U225), transcriptional pausing is likely important for the folding of both domains (19,20), consistent with the inability of pausing-deficient E. coli RNA polymerase to support riboswitch regulation (19). The importance of the kissing loop interaction is also supported by recently solved crystal structures showing that, although cobalamin riboswitches from different organisms differ markedly in their peripheral regions and ligand binding specificity (8,9), the underlying ligand binding mechanism is common to all of them and relies on a kissing loop interaction.
In addition to the kissing loop interaction, the E. coli btuB riboswitch contains additional helical domains (P13 and P14) that participate in the selective sequestration of the RBS sequence (Fig. 1A). Despite available crystal structures and biochemical probing data (8 -10, 17, 18), a clear mechanism describing the relationship between the kissing loop interaction and RBS control still remains to be elucidated. In this study, we have characterized the role of the kissing loop in ligand-induced conformational changes occurring in both the aptamer domain and riboswitch expression platform. Because the kissing loop is located at the interface of the aptamer and expression platform, we speculated that it would be a key element coordinating the aptamer-bound state to riboswitch conformational changes.
Here we report that kissing loop mutations strongly perturb AdoCbl-dependent gene repression in vivo, consistent with the kissing loop being a key element for btuB regulation. By structurally probing nascent riboswitches, we found that disrupting the kissing loop structure effectively uncouples ligand binding from conformational changes occurring in the riboswitch expression platform. These results suggest that, although ligand binding can be achieved using only the aptamer domain, kissing loop formation further stabilizes the ligand-bound state of the btuB riboswitch. The kissing loop was found to not be the only factor influencing riboswitch conformational changes because the comparison of truncated btuB riboswitches revealed that the stability of the P13 helical domain is also of crucial importance. Our work suggests that, whereas the kissing loop is important for ligand binding, it is also implicated as a FIGURE 1. A, schematic representing the secondary structure of the btuB riboswitch when bound to AdoCbl. The RBS is shown as interacting with the A-RBS to form the sequestering stem, thereby inhibiting translation initiation (indicated in red). The AЈ-RBS is shown in orange. The KL tertiary interaction involving P5 and P13 helices is indicated. Paired helical domains (P1-P14) are indicated for each stem and are based on recent phylogenetic analysis (33). B, schematic representing the expression platform in the absence of AdoCbl (19). The helical interaction involving the AЈ-RBS and A-RBS sequences is shown. The RBS sequence is represented as not interacting with the rest of the riboswitch, consistent with the activation of translation in the absence of AdoCbl. C, detailed representation of the btuB riboswitch secondary structure in the ligand-bound state (16). The AЈ-RBS, A-RBS, RBS, and start codon are indicated using the same color scheme shown in A. The dashed line represents the phosphodiester backbone connection between residues 161 and 162, and the arrows show the strand polarity.
transmitting module between the aptamer domain and the expression platform to ensure efficient genetic control.

Experimental Procedures
Bacterial Strains-Bacterial strains used in this study were derived from E. coli MG1655. All btuB mutations were generated as described previously (21). Translational BtuB-LacZ fusions (Table 1) were made using a PCR procedure (22,23).
Templates for In Vitro Transcription-Templates were produced using a lacUV5 promoter (24) and amplified from E. coli MG1655 genomic DNA. Kissing loop riboswitch mutants were amplified using bacterial colony PCR from ␤-galactosidase strains ( Table 2). Templates for the expression platform (residues 162-218) were made by using the T7 promoter (25).
Radioactive Labeling of Trinucleotide Initiator-To ensure specific end labeling of the btuB riboswitch in single round transcriptions, a GCC trinucleotide (Oligos, Etc.) was radiolabeled using a polynucleotide kinase, T4 (Thermo Scientific). Briefly, 5 pmol of trinucleotide was labeled according to the manufacturer's protocol using [␥-32 P]ATP at 37°C for 1 h and directly added to the transcription mixture without further purification.
Single Round in Vitro Transcription Assays-btuB transcripts were produced by in vitro transcription using E. coli RNA polymerase (Epicenter). Transcription reactions were prepared using 300 fmol of DNA template, 300 nM CTP and GTP, 1 pmol of 32 P-labeled GCC trinucleotide, and 1ϫ TB (20 mM Tris-HCl, pH 8.0, 20 mM NaCl, 20 mM MgCl 2 , 14 mM ␤-mercaptoethanol, and 0.1 mM EDTA). The transcription mixture was incubated at 37°C for 5 min. A second reaction mixture containing 5 M UTP and 0.2 unit/l E. coli RNA polymerase was then added, and the resulting mixture was incubated at 37°C for 15 min to allow transcription initiation. A third mixture containing 100 M nucleotides, 1ϫ TB, and 1 mg/ml heparin was added to the reaction tube and incubated 15 min at 37°C for transcription elongation in single round conditions. AdoCbl was added to the required concentration. Experiments described in Fig. 8B were performed by further addition of 150 fmol of a purified transcript corresponding to the riboswitch expression platform (162-218 nt). The resulting mixture was incubated at 37°C for 5 min.
RNase H Cleavage Analysis-After transcription reactions were completed, a mixture containing 1ϫ cleavage buffer (5 mM Tris-HCl, pH 8.0, 20 mM MgCl 2 , 100 mM KCl, 50 M EDTA, and 10 mM ␤-mercaptoethanol) was added together with a 20 M concentration of an oligonucleotide probe and incubated for 5 min at 37°C. Then 0.1 unit/l ribonuclease H (Ambion) was added and incubated for 5 min at 37°C. Where indicated, RNase H reactions were performed using 50 M DNA oligonucleotide. Reactions were stopped using an equal volume of formamide loading dye. K switch values were obtained using a procedure similar to that used for T 50 (26,27). The quantification assumes a simple 1:1 stoichiometry between the aptamer and AdoCbl as expected from in-line probing data (10,16). Reported values represent an average of at least three independent experiments.
RNase T1 Cleavage Analysis-Transcription reactions were performed on a solid surface using a 5Ј-biotinylated DNA template (28). Reactions were prepared using 2 pmol of DNA template, 1ϫ TB1 (5 mM Tris-HCl, pH 8.0, 100 mM KCl, and 10 mM MgCl 2 ), 1 pmol of RNA polymerase ␤Ј-His tag, and 2 pmol of 70 factor and incubated for 5 min at 37°C. A solution of 10 M GCC trinucleotide, 25 M CTP, 25 M GTP, and 15 Ci of [␣-32 P]UTP was then added and incubated for 8 min at 37°C. A mixture of 10 l of magnetic avidin beads was added to the reaction and incubated for 10 min at 37°C. The resulting transcription complexes attached to the beads were washed twice with a solution of 5 mM Tris-HCl, pH 8.0, 1 M KCl, and 10 mM MgCl 2 and resuspended in TB. Transcriptions were completed by the addition of 100 M NTPs and 1 mg/ml heparin in the absence or presence of AdoCbl and incubated for 15 min at 37°C. The supernatant was subjected to RNase T1 (0.2 unit/l) digestion for 1 min at 37°C. Reactions were stopped using phenol-chloroform-isoamyl alcohol extraction and mixed with an equal volume of formamide loading dye. Ladders were generated using 3Ј-O-methyl nucleotides as described previously (27).
Preparation of Transcripts Corresponding to the Expression Platform-A btuB transcript corresponding to the expression platform (positions 162-218) was produced by T7 RNA polymerase transcription. A GCG sequence was added at the beginning of the transcript to minimize the 5Ј heterogeneity of the RNA population (29). Transcription and RNA purifications steps were performed as described previously (30).
Selective 2Ј-Hydroxyl Acylation Analyzed by Primer Extension (SHAPE) Analysis-SHAPE was performed as described previously (31). RNA samples were slowly cooled in a solution containing 100 mM K-HEPES, pH 8.0, and 100 mM NaCl.  Where indicated, reactions were performed in the presence of 10 mM MgCl 2 and 20 or 50 M AdoCbl. N-Methylisatoic anhydride dissolved in dimethyl sulfoxide was added and incubated for 80 min at 37°C. N-Methylisatoic anhydride-treated RNA was precipitated, resuspended, and reverse transcribed using the oligonucleotide BtuB202 (Table 3) as described previously (31). ␤-Galactosidase Assays-␤-Galactosidase experiments were performed as described previously (32). Briefly, the bacterial culture was grown overnight in M63 0.2% glycerol minimal medium and diluted 50ϫ into fresh medium. Arabinose (0.1%) was added to induce expression of lacZ constructs, and AdoCbl (5 M) was added where indicated. ␤-Galactosidase activities were calculated as described previously (32) and were relativized to the specific activity of the wild-type construct obtained in the absence of ligand. Reported activities represent data from at least three independent experimental trials.

Results
The Identity of the P1 Stem Is Crucial for btuB Riboswitch in Vivo Regulation-In most riboswitches, the P1 stem is directly involved in regulating gene expression through riboswitch conformational changes (33). We have recently shown for Bacillus subtilis riboswitches responding to adenine, lysine, and S-adenosylmethionine (27,34,35) and for the E. coli lysine-sensing riboswitch (32) that ligand-dependent gene regulation is highly dependent on the formation of P1 stem base pairs but not on the identity of residues involved. Because riboswitch crystal structures show that AdoCbl sensing is achieved through kissing loop tertiary interaction rather than using the P1 stem (8,9), we investigated the role of the P1 stem in gene regulation by engineering riboswitches with various P1 stem mutations designed to disrupt or allow the formation of P1 stem base pairs ( Fig. 2A). We created BtuB-LacZ translational constructs containing the riboswitch domain with the first 70 btuB codons and assayed riboswitch activity in a defined growth medium in the absence or presence of AdoCbl. The riboswitch construct was fused to an arabinose-inducible promoter to avoid any promoter-specific genetic regulation (24). When assaying the ␤-galactosidase activity of the wild-type riboswitch in the presence of 5 M AdoCbl, the enzymatic activity was decreased by ϳ15-fold compared with the absence of ligand (Fig. 2B). These results suggest that the btuB riboswitch represses gene expression in the presence of AdoCbl in agreement with previous results (10,11,14). To establish the importance of P1 stem formation, we introduced destabilizing P1 stem mutations on either the 5Ј (M1.1) or 3Ј side (M1.2) of the P1 stem ( Fig. 2A). The presence of AdoCbl failed to reduce ␤-galactosidase activity in both cases (Fig. 2B). The simultaneous introduction of M1.1/M1.2 complementary mutations did not allow an efficient AdoCbl regulation (ϳ1.6-fold) (Fig. 2B) in agreement with the genetic control being perturbed compared with the wild-type construct. Given that the P1 stem contains three base pairs exhibiting strict phylogenetic conservation at the sequence level (33), we engineered a second set of riboswitch mutants in which only non-conserved base pairs were altered (M1.3 and M1.4) ( Fig.  2A). Similar to what was observed for M1 and M2 mutants, a complete loss of ligand-dependent genetic regulation was also obtained (Fig. 2B). Moreover, a construct combining both complementary mutations (M1.3/M1.4) resulted in a ligand-induced regulation of ϳ2.9-fold (Fig. 2B), indicating that regulation is not as efficient as in the wild type (ϳ15-fold). Together, our results suggest that the nucleotide sequence of the P1 stem is important for AdoCbl-induced btuB riboswitch regulation. Furthermore, the low ␤-galactosidase activities observed for all tested mutants suggest that the P1 stem sequence is important for btuB gene expression when the concentration of AdoCbl is low.
The Kissing Loop Interaction Is Required for btuB Riboswitch Control-According to cobalamin riboswitch crystal structures (8,9), the kissing loop interaction is in proximity to the bound metabolite, suggesting that the interaction is important for gene regulation. This is consistent with mutations of the kissing loop interaction resulting in a complete loss of AdoCbl-dependent control and complementary mutations only partially restoring regulation (10). However, because crystal structures do not show extensive interaction between the bound ligand and the kissing loop (8,9), it indicates that the identity of residues is not crucial for ligand binding. Thus, to determine the role of kissing loop residues in genetic regulation, we introduced several mutations in the kissing loop interaction and assessed their effect on riboswitch regulation using ␤-galactosidase assays. When introducing 187C3 G/188G3 C mutations inhibiting the formation of two kissing loop base pairs (K1 mutant; Fig. 2C), ␤-galactosidase activity increased both in the absence and presence of AdoCbl (Fig. 2D), indicating that the btuB riboswitch control is strongly perturbed. The introduction of compensatory mutations restoring the formation of both base pairs (K1c; Fig. 2C) resulted in significantly decreased btuB expression regardless of the presence of ligand (Fig. 2D). This observation is consistent with previous results (19). A small degree of ligand-induced regulation (ϳ1.7-fold) was obtained for the K1c mutant, showing that the regulation is altered compared with the wild type (Fig. 2D). A similar analysis was also performed for other base pairs of the kissing loop (K2 and K3 mutants; Fig. 2C). In each case, the presence of AdoCbl failed to modulate the ␤-galactosidase activity (Fig. 2D), consistent with the importance of these base pairs for btuB riboswitch regulation. Compensatory mutations restored riboswitch control only to a limited extent because AdoCbl-dependent repression of ϳ1.2and ϳ1.9-fold was obtained for K2c and K3c mutants, respectively (Fig. 2D). Thus, our results suggest that the identity of kissing loop residues is important for riboswitch gene regulation. However, in contrast to the P1 stem (Fig. 2B), kissing loop mutations did not significantly lower ␤-galactosidase activities (Ͻ0.25 unit) in the absence of ligand, suggesting that both structures play different roles in btuB regulation. Structural Probing of the btuB Riboswitch-To gain insight into AdoCbl-induced btuB structural changes, we performed SHAPE analysis using the electrophilic reagent N-methylisatoic anhydride, which reacts with 2Ј-OH groups located in RNA flexible regions (36). In this assay, purified riboswitch transcripts were renatured and analyzed for their ability to undergo structural changes in the presence of ligand. SHAPE analysis showed that increasing the AdoCbl concentration resulted in several btuB riboswitch conformational changes (Fig. 3A), indicating that ligand binding is effective under our experimental conditions. As reported previously (9), AdoCbl-dependent protection was observed in regions corresponding to stems P10 -P13 and in the single-stranded regions J1/13 and J7/8 (Fig. 3A). However, a weak ligand-dependent protection was obtained in the kissing loop region (C50 residue; Fig. 3, A and B), suggesting an inefficient ligand-dependent folding of this region. Given that the folding of the btuB riboswitch expression platform was recently reported to be highly dependent on the transcriptional context, (19), we sought to investigate AdoCbl-induced riboswitch conformational changes on nascent transcripts.
The RNase H cleavage assay is a well established technique whereby a DNA oligonucleotide is used to target an RNA sequence that is cleaved upon RNase H recognition (19,37,38).
The native structure of the btuB riboswitch was previously investigated using a combination of in vitro single round transcription and RNase H assays (19). In that study, nascent btuB  riboswitches transcribed by E. coli RNA polymerase were probed in vitro with a series of DNA oligonucleotides covering the entire riboswitch (19). Thus, to characterize the influence of the kissing loop interaction on btuB riboswitch structural changes, we used a similar approach in which three DNA oligonucleotides were used to target the aptamer, kissing loop, and P13 stem (Fig. 4A and Table 3). All oligonucleotides were selected based on their ability to report riboswitch conformational changes upon AdoCbl binding according to SHAPE (Fig.  3) and RNase H data (19).
Single round transcription assays were performed using a DNA construct comprising the lacUV5 promoter and the btuB riboswitch fused to the first 70 btuB codons. In the absence of RNase H, transcription reactions yielded a major product corresponding to the full-length transcript (Fig. 4B, lane N). However, when RNase H cleavage assays were performed using a DNA probe targeting the aptamer domain, nascent riboswitch RNAs were cleaved at low AdoCbl concentrations and generated a specific cleavage product (Fig. 4B, product P). Furthermore, when experiments were performed over a range of AdoCbl concentrations, protection from RNase H cleavage was observed (Fig. 4B), consistent with the formation of an AdoCblriboswitch complex resistant to RNase H cleavage as observed previously (19). As a result, we quantified RNase H cleavage products and fit the data to a two-state binding model (27), allowing us to determine the AdoCbl concentration at which half of the riboswitch molecules are in the off state (K switch ). We obtained a value of 55 Ϯ 3 nM, which is approximately half of that reported previously (ϳ93 nM) (19). The presence of cleavage doublets suggests that RNase H cleavage results in the formation of a second product as observed previously in some cases (19,39,40). We obtained a similar value when using a DNA probe concentration 2.5 times greater (K switch ϭ 52 Ϯ 1 nM; Fig. 4C), indicating that the probe was used at saturation. The K switch value is lower than the dissociation constant of the AdoCbl-aptamer complex determined using renatured RNAs (ϳ300 nM) (10), most probably reflecting the importance of native RNA structures in the regulation of the btuB riboswitch (19).
To establish the effect of AdoCbl binding on the formation of the kissing loop interaction, we performed RNase H cleavage assays using a DNA probe that targeted the kissing loop structure (KL probe; Fig. 4A). As observed for the aptamer probe, we detected an efficient RNase H cleavage activity in low AdoCbl concentrations that was progressively reduced when transcriptions were performed at higher ligand concentrations (Fig. 4D). Fitting analysis yielded a K switch value of 122 Ϯ 14 nM, which was higher than the value obtained using the aptamer probe (55 Ϯ 3 nM). This degree of variation between K switch values is expected because they are likely affected by the ability of a given DNA probe to hybridize to its RNA sequence, which can differ markedly due to the relative stability of targeted regions. Interestingly, the ligand-dependent protection of the kissing loop structure suggests that ligand binding stabilizes the formation of the kissing loop, consistent with crystal structures showing that bound AdoCbl is located close to the kissing loop structure (8,9). We also investigated the riboswitch expression platform by targeting the P13 stem (P13 probe; Fig. 4A). As observed for both aptamer and kissing loop probes, RNase H cleavage assays showed that the P13 helical domain is protected in the presence of AdoCbl (K switch ϭ 219 Ϯ 9 nM; Fig. 4E), consistent with the increased stability of the P13 stem in the AdoCbl-riboswitch complex. Together, our results show that structural changes occur in three different regions of the btuB riboswitch in the presence of AdoCbl, consistent with the formation of a more stable structure resistant to RNase H cleavage activity. Furthermore, even though K switch values vary depending on the DNA oligonucleotides used, they provide a powerful way to probe specific regions of the btuB riboswitch, thereby allowing characterization of ligand-induced conformational changes of nascent riboswitch transcripts.
The Kissing Loop Is Crucial for Coordinating Ligand Binding and Riboswitch Conformational Changes-According to in vivo data obtained using translational BtuB-LacZ constructs (Fig.  2D) (10), the K1 kissing loop mutant strongly perturbs liganddependent riboswitch regulatory control. To characterize the role of the kissing loop structure in btuB riboswitch regulation, we used RNase H cleavage assays to investigate the structure of the K1 riboswitch mutant and its ability to perform ligandinduced structural changes. When using the aptamer probe, we determined that AdoCbl-dependent protection could be achieved only at very high ligand concentrations (Fig. 5A), yielding a K switch value of 12,060 Ϯ 2,700 nM. This high value suggests that, although the K1 mutant riboswitch can still bind AdoCbl at very high concentrations, the destabilization of the kissing loop interaction clearly affects ligand binding activity. We also performed RNase H assays using a riboswitch variant containing the mutations 50C3 G/51G3 C, which destabilize the same base pairs as in the K1 mutant but through the corresponding pairing partners (K1.1 mutant; Fig. 2C). A result similar to that of the K1 riboswitch was obtained (K switch ϭ 9,998 Ϯ 2,700 nM), consistent with the importance of these residues for AdoCbl binding. However, when both K1 and K1.1 mutations were simultaneously introduced to allow base pair formation (K1c mutant; Fig. 2C), the ligand-induced protection from RNase H cleavage was observed at significantly lower AdoCbl concentrations (Fig. 5A). The K switch value (369 Ϯ 33 nM) was found to be only ϳ7-fold higher than the wild type, indicating that kissing loop base pairs C50-G188 and G51-C187 are important for forming a stable AdoCbl-riboswitch complex. A similar fitting analysis using the P13 stem probe data yields K switch values for the wild type (219 Ϯ 9 nM) and K1c mutant (329 Ϯ 23 nM). Note that the absence of significant variation does not allow a K switch determination for the K1 mutant when using kissing loop (B) or P13 stem (C) probe. Data for the wild-type riboswitch using the kissing loop (B) and P13 stem (C) probes are taken from Fig. 4, D and E, respectively, and are shown for comparison. Experiments were done in triplicate, and the data shown are a representative result. N.A., not applicable.
Kissing loop riboswitch mutants were also characterized using the probe targeting the kissing loop structure. To assess the formation of the kissing loop in the context of the K1 mutant, we used a different DNA probe (KLc probe; Table 3) that matches the mutations introduced to the L13 loop. When RNase H cleavage assays were performed with the K1 mutant, AdoCbl could not modulate the relative cleavage even at high concentrations (Fig. 5B). These findings indicate that the kissing loop interaction is not stabilized in the context of the K1 mutant. However, when using the K1c riboswitch variant in which the kissing loop is allowed to form, ligand-dependent protection of RNase H cleavage activity was recovered and produced a K switch value of 275 Ϯ 21 nM (Fig. 5B). This demonstrated the formation of the kissing loop interaction in the presence of compatible Watson-Crick pairing residues. However, because this K switch value is lower than that determined for the aptamer domain (369 Ϯ 33 nM), which is in contrast to the wild type, it suggests that the presence of the K1c mutations perturbs the ligand-dependent stabilization of the riboswitch.
We also performed RNase H cleavage assays using the P13 probe to characterize the influence of the kissing loop on conformational changes taking place in the expression platform. As observed for the kissing loop probe, AdoCbl titration completely failed to reduce RNase H cleavage activity when testing the K1 mutant (Fig. 5C), showing the inability of the expression platform to undergo AdoCbl-dependent structural changes. However, a significant decrease in RNase H cleavage was detected using the K1c mutant (K switch ϭ 329 Ϯ 23 nM; Fig. 5C), consistent with the importance of a functional kissing loop structure for inducing conformational changes in the expression platform.
Together, these results clearly show that the presence of the kissing loop is important, but not essential, for btuB riboswitch binding AdoCbl in a transcriptional context. However, our data also indicate that the kissing loop structure is crucial for coordinating aptamer ligand binding activity to the btuB riboswitch conformational changes taking place in the expression platform.
Probing the btuB Riboswitch Native Structure Using RNase T1 Assays-To seek additional information about AdoCbl-induced structural changes occurring in btuB nascent transcripts, we used a biotin-based, solid-phase transcription approach to probe riboswitches using partial RNase T1 digestion (41). By performing the assays in the absence or presence of AdoCbl, several ligand-dependent changes were detected throughout the riboswitch structure, notably in stems P8 -P11 and in regions J1/13 and J3/6 (Fig. 6A). RNase T1 protection was also observed for residues G188 and G190, which are involved in the kissing loop (Fig. 6A). No structural changes were observed in the A-RBS or RBS regions. By monitoring the accessibility of the J1/13 region over a range of ligand concentrations, a K switch of 220 Ϯ 59 nM was obtained (Fig.  6B), which is similar to values determined using RNase H probes targeting the kissing loop or P13 structures (Fig. 4, D  and E). RNase T1 probing was also performed on the K1 mutant (Fig. 6A). Notably, the high accessibility of kissing loop residues G188 and G190 in the K1 mutant agreed with the high RNase H cleavage activity obtained using the KL probe (Fig. 5B). Thus, these results support the RNase H data showing that AdoCbl stabilizes the aptamer and kissing loop structures in the btuB riboswitch. The Stability of the AdoCbl-Riboswitch Complex Is Improved through the Kissing Loop Interaction-It was recently reported that three RNA polymerase pause sites are located in the btuB riboswitch expression platform (19). One pause site was mapped in the P13 stem at position C184 (Fig. 1C). Because ϳ12 nucleotides are present in the RNA polymerase exit channel (41,42), a transcriptional complex paused at C184 would likely allow about 172 nucleotides of the btuB riboswitch to fold, thereby allowing aptamer domain formation. However, the kissing loop interaction would not take place because this sequence would not yet have been transcribed, thereby potentially precluding ligand binding. To investigate whether the aptamer domain can perform AdoCbl binding, we engineered a truncated version of the btuB riboswitch containing only the first 161 nucleotides, effectively corresponding to the aptamer domain (Fig. 7A). Using this construct, we performed RNase H cleavage assays while varying the AdoCbl concentration (Fig.  7B). Protection from RNase H cleavage occurred only at high ligand concentrations of the aptamer probe, indicating that the aptamer domain can bind ligand although with a drastically reduced efficiency (K switch ϭ 62,310 Ϯ 2,263 nM). This finding is consistent with the importance of the kissing loop for AdoCbl binding (Fig. 5A).
An additional pause site was found at position U225 (19), suggesting that nascent btuB riboswitches in paused transcription complexes expose ϳ213 nucleotides, allowing both the kissing loop and P13 helical domain to form. We engineered a transcription construct in which the riboswitch domain was truncated at position 202 and used RNase H assays to assess the ability of this sequence to perform AdoCbl sensing. This particular RNA size was chosen because it is typically used for in-line probing experiments to study the binding features of the btuB aptamer (10, 16 -18). The ligand-induced stabilization of the AdoCbl-riboswitch complex was significantly improved when using the 202-nt construct compared with the shorter 161-nt construct (K switch ϭ 234 Ϯ 26 nM; Fig. 7B). Because the btuB secondary structure model suggests that the P13 stem contains at least four more base pairs (Fig. 1C), we also used a construct truncated at position 206, enabling complete formation of the predicted P13 stem. RNase H cleavage assays revealed that these btuB riboswitch transcripts exhibit a ϳ2-fold improvement in ligand-dependent protection from RNase H cleavage (K switch ϭ 112 Ϯ 31 nM; Fig. 7B). The influence of P13 stability on the AdoCbl-dependent formation of the kissing loop was also assayed with kissing loop and P13 stem DNA probes (Fig. 7, C and D). In these experiments, the 161-nt construct was not tested due to the absence of the kissing loop interaction. RNase H assays performed on the 202-and 206-nt constructs showed that ligand-induced formation of the kissing loop structure is highly dependent on the stability of the P13 helical domain given that K switch values of 3,929 Ϯ 276 and 158 Ϯ 72 nM were obtained for the 202-and 206-nt constructs, respectively (Fig. 7C). As expected, we also found that a stable P13 stem is needed for AdoCbl to induce conformational changes in the expression platform because RNase H cleavage assays using the P13 probe showed that the 206-nt construct yielded a K switch value of 407 Ϯ 23 nM (Fig. 7D). No K switch value could be determined for the 202-nt construct due to the absence of significant change. Taken together, our results suggest that although ligand binding can be achieved at the first transcriptional pause site (C184), the stability of the RNA-ligand complex is strongly increased at the second pause site (U225) when the P13 stem is transcribed.
The P13 Stem Is Important for the Adoption of the Ligand-free Conformation-Based on predicted btuB structures (Fig. 8A) (19), ligand binding to the riboswitch rearranges the P13 stem to allow the kissing loop interaction. The AЈ-RBS sequence appears to be the dominant feature that controls gene expression by selectively sequestering the RBS in the presence of AdoCbl (Fig. 8A). However, because a structural reorganization of the P13 stem has been proposed based on phylogenetic analysis (19), it suggests that the kissing loop interaction has a role in controlling access to the RBS region (Fig. 8A).
To decipher the role of the P13 stem in riboswitch regulation, we engineered btuB riboswitch constructs harboring mutations in P13 and P14 helical domains. We first disrupted the stability of the P13 stem by substituting residues 173-176 with their Watson-Crick complement (M13.1; Fig. 8A). The inclusion of C173 and U174 mutations in M13.1 was done to prevent their putative involvement with J13/14 residues in the formation of the P13 stem. When assaying the effect of these mutations in a translational BtuB-LacZ construct, the ␤-galactosidase activity in the absence and presence of AdoCbl was found to be dramatically decreased (Fig. 8B), indicating that the P13.1 mutant is constitutively repressed. We also mutated base pairs U187-G198 and G188-U197, which are predicted to form only in the absence of AdoCbl (M13.2; Fig. 8A). When changing the identity of these nucleotides for their Watson-Crick complement, very low btuB expression was observed both in the absence and presence of ligand (Fig. 8B). We obtained similar results by introducing 200U3 A/201G3 A mutations (M13.3; Fig. 8B), indicating that P13 stability is crucial for btuB regulation. As a control, we altered the sequence in the P14 loop that is not predicted to be involved in btuB gene regulation (M14.1; Fig.  8A). By introducing Watson-Crick complementary residues in the region 219 -223, we obtained a mutant that exhibited a decreased (3.5-fold) regulation efficiency of ligand-dependent repression (Fig. 8B), indicating that the riboswitch mutant can perform AdoCbl-dependent regulation. These results clearly indicate that formation of the P13 stem not only stabilizes the AdoCbl-riboswitch complex (Fig. 5B) but also promotes btuB expression in the absence of ligand by maintaining the A-RBS/ AЈ-RBS interaction (Fig. 8A).

Discussion
RNA-based gene regulation mechanisms inherently rely on the correct folding of intervening RNA molecules, ranging from simple productive base pairing to complex tertiary structure formation (2,5,(43)(44)(45). As found for large RNA molecules such as the group I intron and RNase P (45), the transcription process is an integral part of the RNA folding pathway because the polarity of transcription allows incremental availability of newly synthesized RNA sequences, therefore modulating the formation of upstream RNA structure. This polarity-driven folding process is further controlled through the strategic location of transcriptional pause sites (46) that presumably allow sufficient time for secondary and tertiary interactions to take place. These interactions are often important for the folding of complex RNA structures (47,48). Transcriptional pausing has been identified in metabolite-binding riboswitches responsive to flavin mononucleotide (49), adenine (27), and AdoCbl (19) and in a pH-responsive riboregulator (50).
Because the btuB riboswitch relies on the transcription process to achieve its native folding (19), structural studies characterizing btuB folding should preferentially be done using nascent transcripts. However, the low transcription efficiency of E. coli RNA polymerase in vitro, compared with T7 RNA polymerase, creates a more technically challenging problem in probing RNA structure. Conventional probing techniques (e.g. SHAPE analysis) often require large amounts of RNA to detect chemical modifications through reverse transcription. Although SHAPE analysis allows us to monitor btuB structural changes upon ligand binding (Fig. 3) (9), it has been shown that transcriptional pausing is critical for coordinating the ligand binding and riboswitch conformational changes taking place in the expression platform, thus requiring the use of nascent transcripts. RNase H cleavage assays are well suited to study nascent transcripts as they require much less RNA and can be specifically targeted to different transcript domains. However, in contrast to other probing techniques that directly assess ligand binding, RNase H cleavage activity relies on the hybridization of an oligonucleotide that potentially introduces a bias into the study. In the present case, K switch values obtained by targeting different riboswitch domains range from ϳ50 to ϳ200 nM, which are in the range of the reported 300 nM K d obtained using in-line probing assays (10) and RNase T1 analysis (Fig. 6B). Thus, these values suggest that oligonucleotide binding does not significantly perturb the formation of the riboswitch-ligand complex. Because variations between K switch values are observed when different oligonucleotides are used, it may be that not only the ligand binding affinity is reflected in the determined K switch value. Nevertheless, the ability to directly assess defined riboswitch structural features coupled to mutagenesis is a powerful approach to comprehend the riboswitch folding mechanism.
In this study, we further characterized the folding of the cobalamin-sensing btuB riboswitch, which is one of the largest known riboswitches exhibiting complex secondary and tertiary structure (8,9,15,16). Such an intricate RNA architecture is presumably important to specifically recognize the relatively large AdoCbl metabolite, which is mainly achieved through shape complementarity (8,9). Combined with previous findings (19), our results support a model in which transcriptional pausing at C184 allows time for btuB aptamer domain formation and where the presence of AdoCbl results in the sequestering of the AЈ-RBS region via the formation of the P1 stem. Although such a paused complex contains all necessary elements to bind AdoCbl, RNase H data suggest that the free concentration of AdoCbl would need to be in the micromolar range to generate the riboswitch-ligand complex (Fig. 7B). However, our results indicate that further synthesis of the btuB riboswitch leading to the second transcriptional pause site (U225) significantly improves AdoCbl affinity. The dominant structural feature contained in the newly transcribed btuB portion corresponds to the kissing loop interaction, therefore implying that the formation of the kissing loop is important for AdoCbl binding in agreement with its close proximity to the ligand in crystal structures (8,9). Because RNase H data show that btuB transcripts containing 202 and 206 nucleotides, respectively, exhibit a 265-and a 554-fold decrease in K switch values when compared with the 161-nucleotide transcript (Fig. 7B), both kissing loop formation and P13 stability seem to be crucial elements in AdoCbl recognition. Because it was previously reported that the kissing loop can be formed in the absence of AdoCbl at elevated Mg 2ϩ concentrations (9), it suggests that the kissing loop exhibits structural dynamics allowing it to sense the presence of AdoCbl, which is consistent with a conformational capture mechanisms as observed previously for other riboswitches (7,35,51).
Our results indicate that the kissing loop interaction is also important for coordinating AdoCbl binding and riboswitch conformational changes taking place in the expression platform. This is apparent from the complete absence of RNase H cleavage protection observed for the K1 mutant when using KL and P13 probes (Fig. 5, B and C). Such an unconventional role for a kissing loop interaction differs from previous studies reporting that kissing loop structures are usually involved in stabilizing the ligand-riboswitch complex as found for purine and lysine riboswitches (34,52). The presence of multiple transcriptional pause sites located across the expression platform indicates that the folding process is highly controlled in this region, consistent with the decreased folding response in the presence of ligand when the riboswitch is transcribed using T7 RNA polymerase (19). Because of such strategically located pause sites, it is likely that the formation of the kissing loop structure directs the folding of the P13 and P14 helical domains that are involved in riboswitch regulation, consistent with RNase H data (Fig. 5B). This regulatory control involves a "helix sliding" mechanism that rearranges some of the P13 base pairs (Fig. 1A) (19), which is presumably important for P14 folding. Cobalamin-responsive riboswitches controlling translation typically exhibit L13 loop elements containing the RBS sequence, thus providing a direct mechanism for modulating translation initiation through kissing loop formation (8,9). This is in contrast to the E. coli btuB riboswitch containing P13 and P14 structural elements (Fig. 1A), the latter being directly implicated in RBS-selective sequestration. Secondary structure predictions suggest that ligand binding to the btuB riboswitch results in a major rearrangement of the P13 helical domain in which L13 residues become available for kissing loop interaction (Fig. 8A). Surprisingly, ␤-galactosidase assays show that destabilizing the P13 base pairs (M13 mutants) severely limits AdoCbl-induced gene regulation by markedly repressing gene expression (Fig. 8B). In particular, mutations destabilizing base pairs C187-G198 and G188-U197 (M13.2 mutant) strongly alter AdoCbl-dependent gene repression (Fig. 8B), indicating that the stability of P13 base pairs comprising kissing loop residues is important for riboswitch activity. These results imply that P13 stability is important mostly for maintaining btuB expression in the ligand-free form. Precedence for such a helical conformational change has been observed for the Varkud satellite ribozyme where an intramolecular kissing loop interaction induces a helical rearrangement of the substrate stemloop required for ribozyme catalysis (53). Although this mechanism presents characteristics similar to that of the btuB riboswitch, there is still no direct evidence indicating whether P13 conformational changes are performed as a consequence of P1 stem formation, through the interaction of the kissing loop, or both. However, half-life measurements performed in the absence or presence of AdoCbl indicate that ligand binding to btuB mRNA can occur post-transcriptionally (11), suggesting that pre-existing riboswitches can undergo conformational rearrangement to repress gene expression upon ligand binding. Clearly, additional work will need to be done to understand the molecular basis of the structural switching.
Similar to M13 mutants, sequence changes introduced in the P1 stem perturb btuB genetic control by strongly repressing expression regardless of AdoCbl presence (Fig. 2B). These findings are in contrast to other riboswitches where mutations introduced in the 5Ј side of the P1 stem usually result in the derepression of gene expression (27,32). Given that M1.2 and M1.4 mutants directly weaken base pair formation of the AЈ-RBS/A-RBS interaction, the A-RBS/RBS helical domain (P14) is readily favored in such a context, thereby resulting in low ␤-galactosidase activities. However, for M1.1 and M1.3 mutants, no obvious deleterious effect is observed because the current regulation model does not take into account the 5Ј side of the P1 stem (Fig. 1A). Interestingly, a secondary structure analysis performed using the program mfold predicts an interaction between the 5Ј side of the P1 stem and L5 residues (48 -52 nt) in the absence of AdoCbl. Thus, in the context of M1.1 and M1.3 mutants, it is possible that the P1/L5 interaction is disabled, which would, therefore, promote the kissing loop interaction and btuB gene repression.
In summary, our study shows that the formation of the long range kissing loop interaction is important for efficient ligand binding and is crucial for the coordination of conformational changes in the riboswitch expression platform. Given that btuB transcriptional pause sites are strategically located to allow formation of key riboswitch structural elements such as the kissing loop (19), transcription elongation is likely a highly controlled process in which discontinuous movements of the RNA polymerase are an integral part of riboswitch gene regulatory control. The presence of transcriptional pause sites in a wide range