Single-molecule FRET observes opposing effects of urea and TMAO on structurally similar meso- and thermophilic riboswitch RNAs

Abstract Bacteria live in a broad range of environmental temperatures that require adaptations of their RNA sequences to maintain function. Riboswitches are regulatory RNAs that change conformation upon typically binding metabolite ligands to control bacterial gene expression. The paradigmatic small class-I preQ1 riboswitches from the mesophile Bacillus subtilis (Bsu) and the thermophile Thermoanaerobacter tengcongensis (Tte) adopt similar pseudoknot structures when bound to preQ1. Here, we use UV-melting analysis combined with single-molecule detected chemical denaturation by urea to compare the thermodynamic and kinetic folding properties of the two riboswitches, and the urea-countering effects of trimethylamine N-oxide (TMAO). Our results show that, first, the Tte riboswitch is more thermotolerant than the Bsu riboswitch, despite only subtle sequence differences. Second, using single-molecule FRET, we find that urea destabilizes the folded pseudoknot structure of both riboswitches, yet has a lower impact on the unfolding kinetics of the thermodynamically less stable Bsu riboswitch. Third, our analysis shows that TMAO counteracts urea denaturation and promotes folding of both the riboswitches, albeit with a smaller effect on the more stable Tte riboswitch. Together, these findings elucidate how subtle sequence adaptations in a thermophilic bacterium can stabilize a common RNA structure when a new ecological niche is conquered.

Previous comparative studies of the isolated Bsu and Tte riboswitches ( formally, their aptamer domains ) have demonstrated that both RNAs adopt similar conformational ensembles under near-physiological buffer conditions with 1 mM Mg 2+ ( 11 ) .Using single molecule fluorescence resonance energy transfer ( smFRET ) to monitor the conformational dynamics revealed that both riboswitches exhibit an 'undocked' pre-folded conformation wherein the 3 A-rich tail transiently interacts with the P1-L1 stem-loop to form P2 in the 'docked' folded conformation.In the presence of preQ 1 , P2 is further stabilized, and the folded state becomes more favorable ( Figure 1 E ) .Despite these general conformational similarities between the two RNAs, the ligand-free Tte aptamer shows characteristically slower dynamics and a higher population of the docked state than the Bsu aptamer, indicating a generally higher degree of structural compactness ( 11 ) .Computational modeling further indicates that preQ 1 binding to the Bsu and Tte aptamers preferentially occurs through conformational selection and induced fit mechanisms, respectively, suggesting distinct folding pathways ( 11 ) .The two folding pathways can be distinguished by the timing of ligand binding relative to full pseudoknot folding.In the conformational selection pathway, ligand binding to a preformed RNA pocket occurs late, while 3 A-rich tail docking and the formation of P2 occur simultaneously with ligand binding.In contrast, in the induced fit pathway an early ligand-binding event initiates the formation of the remaining tertiary contacts.Such distinctions measured at room temperature correlate with the distinct environments wherein the two bacteria exist; the Tte preQ 1 riboswitch originates from a thermophilic bacterium that grows optimally at a high temperature of 75 • C ( 12 ) and thus would be expected to be more thermostable than the mesophilic Bsu riboswitch with an optimum temperature of 35 • C.However, it is unknown whether these two RNAs exhibit differences in their stabilities to thermal and chemical denaturation, given the high similarity of their sequences and tertiary structures.Furthermore, a more detailed understanding of the differences in structural stability between the two aptamers in relation to their solvent environment is also lacking.
Previous biophysical studies have heavily focused on the influential role of mono-and divalent metal cations in the folding of polyanionic nucleic acids ( 11 ,13-15 ) .Recently, osmolytes have been shown to also impact RNA folding equilibria through their interaction with nucleobases, ribose sugars and phosphate groups (16)(17)(18) .Osmolytes are small watersoluble organic molecules utilized by certain cells to regulate osmotic pressure ( 20 ) .Depending on the chemical nature of an osmolyte, it can either promote or inhibit the folding of nucleic acids and proteins ( 21 ,22 ) .Therefore, while minimizing fluctuations in cellular osmotic pressure, it is also critical for cells to maintain proper intracellular concentrations of both structurally destabilizing and stabilizing osmolytes so that their opposing effects on nucleic acid and protein stabilities are counterbalanced.One such osmolyte, urea, is one of the most widely used denaturants of both proteins and nucleic acids (23)(24)(25) , whereas trimethylamine N -oxide ( TMAO ) is known for its abilities to promote biopolymer folding ( 24 ,26-28 ) and counteract the denaturing effect of urea ( 16 , 18 , 24 , 26 , 29 ) .Due to the opposing effects of urea and TMAO on biomolecule stability, we hypothesize that they can be used in combination to finely adjust the degree of conformational transition.By employing smFRET microscopy, we will be able to extract the energetic and kinetic characteristics of structurally similar Bsu and Tte riboswitch aptamers.
In this study, we investigate the thermodynamic stability of the two riboswitch RNAs by directly probing their heat denaturation, which is detected by changes in UV absorption.Our results demonstrate that the Tte aptamer exhibits higher thermal stability compared to the Bsu aptamer, and this stability is significantly enhanced by ligand binding .By combining urea-based denaturation and smFRET, we find that urea has a more pronounced impact on the conformational dynamics of the Bsu riboswitch compared to the Tte riboswitch.Furthermore, we observe a strong interplay among urea, TMAO, and Mg 2+ ions, where TMAO and Mg 2+ effectively counteract the denaturing effects of urea on both RNAs.Our data show that TMAO exerts a stronger protective effect against urea-induced unfolding of the Bsu riboswitch, while its influence on the more stable Tte riboswitch is comparatively less prominent.Finally, m -value analyses of the transition kinetics further support that, upon solvent perturbation, the Tte riboswitch displays a higher structural stability than Bsu .We anticipate that monitoring structural perturbations induced by osmolytes using smFRET microscopy has the potential to serve as a generalizable tool for discerning differences in the conformational free energy landscapes of structurally similar RNAs.

RNA preparation for melting curve studies
Bsu and Tte preQ 1 riboswitch aptamers for melting experiments were generated by in vitro transcription using T7 RNA polymerase ( RNAP ) prepared in-house as previously described by He et al .( 30 ) T7 RNAP was expressed in BL21 with N-terminal His-tag, and purified in 250 mM NaCl preventing co-purification of other proteins.Affinity-based purification was achieved by a hybrid batch / gravity-flow procedure.In vitro transcription reactions were designed based on previously described conditions with minor modifications.DNA oligonucleotides from IDT were purified by 20% denaturing urea polyacrylamide gel electrophoresis and excised under UV-shadowing with a 312 nm lamp.The DNA templates ( Tte : 5 -CCC TTG TTT TGT TAA CTG GGG TTA CTG CGA CCC A GG A CC T A T A GT GA G TCG T A T T AA ATT-3 ; 5 -AAT TTA ATA CGA CTC ACT ATA GG-3 , Bsu :

-CCT TAG TTT TTT A T A GAG GGT GTA ACT AGA ACC TCT GCC T A T A GT GA G TCG T A T T AA A TT -3 ; -AA T TT A A T A CGA CTC ACT
A T A GG-3 ) were eluted from the gel, ethanol precipitated and resuspended in nuclease free water.300 μl transcription reactions containing 10 μM T7-Bsu ( or Tte ) duplex template, 120 mM HEPES-KOH ( pH 7.6 at 22 • C ) , 0.01% ( v / v ) Triton X-100, 30 mM MgCl 2 , 7.5 mM of each NTP, 40 mM DTT, 2 mM spermidine trihydrochloride, 0.2 mg / ml T7 RNAP, and 0.01 U / μl inorganic pyrophosphatase ( MP Biomedicals ) were incubated in a circulating water bath at 37 • C for 16-18 hours, and then mixed with an equal volume of 2 × gel loading buffer ( 95% ( v / v ) formamide, 18 mM EDTA, 0.025% [w / v] each of SDS, bromophenol blue, and xylene cyanol ) to stop the reaction.The reaction with loading buffer was heated at 90 • C for 3 min and then snap cooled on ice.The RNA transcripts were gel purified as described for DNA above.

UV-Vis melting studies
Melting experiments were conducted on a Beckman DU® 640B spectrophotometer with a Beckman High Performance Temperature Controller, a Transport accessory, a T m Six-Cell Holder, and 1 cm path length quartz cuvettes ( Beckman Coulter, 523878 ) .Tris-based buffers were replaced with sodium phosphate as buffering salt, and ionic strength was adjusted to match 1 × smFRET buffer.A typical sample contained 0. Samples were refolded by heating at 90 • C for 3 min, at 70 • C for 3 min, at 60 • C for 3 min, and finally allowed to cool to room temperature over 20 min.For experiments done with saturating ligand, 0.45 μM preQ 1 was added to the refolded RNA ( 1.5:1 ligand:RNA final concentration ratio ) .325 μl sample was transferred to each cuvette and tightly stoppered.A 260 was monitored with a 0.5 s read averaging time, and A 340 or A 320 was used for background correction.Temperature was increased at a rate of 1 • C / min between 10 and 22 • C with a reading made every 1 • C, then at a rate of 0.5 • C / min between 22 and 75 • C with a reading made every 0.5 • C, then at a rate of 1 • C / min between 75 and 95 • C with a reading made every 1 • C.Only denaturation transition curves were collected, and each condition was repeated four times.The melting temperature for each apparent transition was determined as described previously from the second derivative of the absorbance versus temperature plot ( 31 ) .

Preparation of labeled RNA for smFRET experiments
The smFRET constructs for Bsu and Tte preQ 1 riboswitch aptamers were prepared as previously described ( 11 ) .Briefly, the RNAs were chemically synthesized by Dharmacon, Inc. ( Fayette, Colorado ) with modifications as follows: 5biotinylated, 3 -DY547 labeled and 5-aminoallyl uridine ( 5NU ) labeled at position U12 ( Tte ) and U13 ( Bsu ) .The RNAs were first deprotected following the manufacturer's instructions and then labeled with Cy5-NHS ester ( GE Healthcare ) .One dye pack was used for labeling one construct.The dye pack was dissolved in 30 μl of DMSO and used to label 3.4 nmol RNA in a total reaction volume of 50 μl containing 0.1 M sodium bicarbonate buffer ( pH 8.7 ) .The mixture was then incubated and tumbled at room temperature in the dark for 4 h.Excess free dye was removed by using a Nap-5 gel filtration column ( GE Healthcare ) .The RNAs were collected, ethanol precipitated and dissolved in autoclaved, deionized water ( 11 ) .The complete sequences for the RNA constructs used in smFRET experiments were Bsu : 5 -biotin-ugcgggA GA GGUUCU AGCU AC ACCCU CUAUAAAAAACUAAGG-3 and Tte : 5 -biotin-ucacCUGG GUCGC AGUAACCCCAGUUAAC AAAACAAGGG-3 .

Single-molecule FRET experiments
We assembled a microfluidic channel in between a clean quartz slide and a glass coverslip as described by previous smFRET studies ( 32 ) .We then coated the quartz slide with biotinylated-BSA followed by streptavidin.The RNA was folded in 1X smFRET buffer ( 50 mM Tris-HCl, 100 mM KCl, pH 7.5 ) in the absence of Mg 2+ , by heating at a low concentration ( 20-30 pM ) at 90 • C for 1 min, snap-cooling on ice for 30 s, then allowing to slowly reach room temperature for over 15 min.100-200 μl of the heat-annealed RNA was flowed into the microfluidic channel and incubated for 5 min.The surface bound streptavidin was able to capture the 5 -biotin on the RNA construct and ensure the surface immobilization of the RNA.The unbound free RNA was washed away using 1 × smFRET buffer ( ±Mg 2+ depending on the experiment ) .Osmolyte titration was performed with or without 1 mM Mg 2+ , as indicated, and in the presence of 100 nM preQ 1 .An oxygen scavenging system ( OSS ) containing 5 mM protocatechuic acid ( PCA ) , 50 nM protocatechuate-3,4dioxygenase ( PCD ) , and 2 mM Trolox ( 6-hydroxy-2,5,7,8tetramethylchroman-2-carboxylic acid ) was introduced into the microfluidic channel serving the purpose of extending the life of fluorophores and reducing fluorophore photoblinking events ( 32 ,33 ) .The experiments were performed on a prismbased total internal reflection fluorescence ( TIRF ) microscope.A 532 nm laser was used to excite DY547, and emission from both DY547 and Cy5 was recorded concurrently using an intensified charge-coupled device camera ( ICCD, I-Pentamax, Princeton Instruments ) at 60 ms time resolution ( 11 ) .We processed raw movie files using IDL ( Research Systems ) to extract smFRET time traces that were analyzed using MATLAB ( The Math Works ) scripts.We manually selected smFRET time traces meeting the following criteria: single-step photobleaching, a trajectory length of at least 100 camera frames before bleaching, a signal-to-noise ratio of > 4:1, and a total fluorescence intensity ( donor + acceptor ) of > 300 ( arbitrary units ) ( 11 ) .I D and I A represent the background corrected intensities of DY547 ( donor ) and Cy5 ( acceptor ) fluorophores, and the FRET efficiency was calculated as E FRET = I A / ( I D + I A ) .FRET efficiencies observed in first 100 frames from each trace were combined to generate FRET distribution histograms using MATLAB.Dynamic time traces were idealized with a twostate model according to Hidden-Markov Modeling ( HMM ) using a segmental k-means algorithm in QuB software.Transition Density Plots ( TDPs ) and Transition Occupancy Density Plots ( TODPs ) were then generated using MATLAB ( 34 ) .A summary of the total number of molecules collected under each experimental for smFRET analysis is provided in Supplementary Table S1.

Rate constant analysis from smFRET data
Dwell times in the undocked and docked states were extracted from all the idealized traces, and the cumulative dwell time distributions were fit with an exponential function of the form y ) to obtain the rate constants k dock and k undock , respectively, as described previously ( 31 ) .Exponential fitting was performed in MATLAB.In cases where the 95% confidence interval for the fitting parameters contained negative values, or when there were too few data points, the double-exponential fit was rejected in favor of a single-exponential fit.Whenever data were better fit with a double-exponential function, more than one underlying process with each a characteristic rate constant appears to be occurring, involving, e.g.formation of an intermediate.To compare our data across conditions, the average rate constant of the double-exponential fits was calculated as the weighted average of the two rate constants ( A 1 k 1 + A 2 k 2 ) .A summary of the number of dynamic traces that contributed to the rate constant determinations for each experimental condition and whether a single-or double-exponential fit was used is provided in Supplementary Table S2.The error associated with the measured rate constants was estimated by bootstrap fitting on 1000 sets of replicate data that were chosen by sampling M traces from traces in original dataset with replacement, where M is equal to the number of traces present in the original dataset.The standard deviation of the resulting 1000 replicate fit-coefficients is reported as the error of measured rate constant ( 31 ) .

m -value analysis of smFRET data
The change in the Gibbs free energy of docking ( G • dock ) for the riboswitch aptamers in the presence of increasing concentrations of osmolyte was determined from the smFRET distribution histograms.After fitting the FRET distribution to the sum of two Gaussians ( the

Results
UV-Vis melting reveals higher thermal stability for the Tte than the Bsu riboswitch Due to its origin from a thermophilic bacterium, the Tte riboswitch is expected to be more thermostable than the mesophilic Bsu riboswitch.However, there is a lack of data providing a side-by-side comparison of their melting temperatures or thermal stabilities.Therefore, to directly assess and compare the relative stabilities of the two riboswitch pseudoknots, we conducted UV-Vis melting studies in the absence or presence of 1 mM Mg 2+ and of a 1.5-fold molar excess of cognate ligand preQ 1 over RNA ( 0.45 μM preQ 1 , considerably above the known ligand binding affinities in the nanomolar range ( 11 ) , and 0.3 μM RNA ) .Notably, under all conditions tested, the temperature required to fully denature the Tte aptamer was found to be approximately 10-20 • C higher than that of Bsu ( Table 1 ) .The highly thermally stable P1 stem of the Tte riboswitch provides the structural basis for the Tte riboswitch to function at elevated environmental temperatures.
In addition to displaying higher thermal stability, the Tte riboswitch aptamer exhibited stepwise melting of P2 and P1 helices, whereas the Bsu riboswitch aptamer displayed a single melting transition.Specifically, in the absence of Mg 2+ and preQ 1 , the Tte riboswitch showed two distinct melting transitions at approximately 40ºC and 66ºC ( Figure 2 , Table 1 ) .These transitions were previously assigned to melting of stem P2 ( and L3 ) and stem-loop P1, respectively ( 31 ) .The presence of 1 mM Mg 2+ increased the melting temperature of P2 to approximately 48ºC, while the melting temperature of P1 remained unaffected.This stabilizing effect of Mg 2+ is in line with the well-known ability of Mg 2+ to enhance the stabil- ity of RNA tertiary interactions and pseudoknot folding ( 11 ) .Recent structural analysis of the Tte riboswitch further underscored the significance of divalent ions in its folding process, as it revealed the presence of six divalent metal ion binding sites within both the apo and ligand-bound structures of Tte ( 35 ) .On the other hand, upon addition of preQ 1 , we observed a significant increase in the melting temperatures of P2 and P1, regardless of the absence and presence of Mg 2+ ( Figure 2 , Table 1 ) .The Tte preQ 1 riboswitch is thought to follow the induced fit pathway, where the preQ 1 ligand binds to the RNA with high affinity, inducing pseudoknot folding ( 11 ,31 ) .Therefore, the folding of Tte is highly dependent on the presence of the ligand, consistent with our experimental findings.Moreover, recent structures of Tte revealed the ligand-induced coaxial stacking of the P2 and P1 stems ( 35 ) .This phenomenon provides an explanation for the ligand-induced increase in the melting temperatures of both P2 and P1, as the coaxial stacking contributes additional stability to the structure.In contrast, for the Bsu riboswitch, we observed only a single melting transition at ∼54ºC, which remained largely unaffected by the presence of Mg 2+ and preQ 1 ( Figure 2 , Table 1 ) .Notably, Bsu does not exhibit a distinct gap in the thermal stability between P1 and P2, as the formation of P2 is an intrinsic characteristic of the RNA structure rather than being induced by the additional stability provided by ligand binding ( 13 ) .
smFRET reveals distinct conformational responses for the Bsu and Tte riboswitches in the presence of urea While the ensemble UV-Vis melting analysis unveils the relative thermodynamic structural stabilities of the Bsu and Tte riboswitches, it lacks information on their folding kinetics.Previously, we used smFRET to study the P2 helix folding kinetics of the Bsu and Tte riboswitches in the presence of Mg 2+ and various ligands that bind and stabilize the folded structure ( 11 ,13 ) .To understand the kinetic origin of the observed differences in RNA stability, we therefore utilized smFRET to investigate the pseudoknot formation ( i.e. docking ) dynam-ics of these riboswitches in the presence of the osmolytes urea and TMAO ( Figure 3 A ) .We kept the preQ 1 concentration at 100 nM for all experiments to enable observation of folding dynamics, a condition where the full P2 stem reversibly docks even in the absence of Mg 2+ .Accordingly, single molecule time trajectories demonstrate two-state folding kinetics with well-defined mid-FRET and high-FRET states in the absence of Mg 2+ ( Figure 3 B; the molecule number for each smFRET experiment presented here-typically between ∼50 and 350-is reported in Table S1 ) , consistent with previous assignments of the undocked and docked pseudoknot conformations, respectively ( 11 ,13 ) .The corresponding FRET histogram for the Bsu riboswitch exhibits a mid-FRET state, representing the undocked conformation ( mean FRET efficiency, μ undock ∼0.57with 61% population weight; Figure 3 C ) , and a high-FRET state representing the docked conformation ( μ dock ∼0.98, 39%; Figure 3 C ) .Notably, the mean FRET efficiency of the mid-FRET state, μ undock , shifts gradually to lower values with increasing concentration of urea to reach μ undock ∼0.4 at 4 M urea, while in parallel the occupancy of the high-FRET state steadily decreases to only 8% with only a small reduction in μ dock to 0.90 ( Figure 4 A, top panel ) .This shows that urea unfolds the docked pseudoknot conformation, as indicated by a decrease in percentage occupancy at μ dock .This is consistent with the ability of urea to denature RNA structures ( 16 ,25 ) .Similarly, FRET histograms for the Tte riboswitch show a mid-FRET state of μ undock ∼0.73 ( 53% ) , a high-FRET state at μ dock ∼0.98 ( 47% ) , and display generally similar trends as the Bsu riboswitch with both mid-FRET values and high-FRET state occupancy decreasing in the presence of increasing urea concentrations ( Figure 4 A, bottom panel ) .However, the value of mid-FRET at 4 M urea remains relatively higher for Tte ( μ undock ∼0.6 ) than that for Bsu ( μ undock ∼0.4 ) , suggesting less opening of the Tte riboswitch upon urea denaturation than the Bsu riboswitch despite their overall similar topology and sequence ( Figure 1 A ) .
From the smFRET traces, dwell times in the mid-and high-FRET states were extracted using a two-state Hidden Markov Model ( HMM ) and fit with single-exponential functions to estimate the rate constants of docking ( k dock ) and undocking ( k undock ) , respectively.For Bsu, in the presence of 100 nM preQ 1 but no Mg 2+ , k dock decreases gradually from 1.35 s −1 in the absence of urea to 0.1 s −1 in the presence of 4 M urea ( Supplementary Figure S1a ) .Conversely, k undock increases from 0.5 s −1 in the absence of urea to 2.3 s −1 in the presence of 4 M urea ( Supplementary Figure S1a ) .A similar trend was observed for the Tte riboswitch; k dock decreases from 4.3 s −1 to 0.5 s −1 and k undock increases from 1.2 s −1 to 11 s −1 from 0 to 4 M urea ( the latter value approaching our 60 ms time resolution; Figure S1b ) .These observations suggest that, in the absence of Mg 2+ , urea destabilizes P2 formation by both slowing docking and accelerating undocking for both the Bsu and Tte riboswitches, similar to prior observations for an idealized tertiary structure docking system ( 18 ) .
Transition-state analysis ( ) is a valuable tool for studying the nature of the transition state, elucidating the origin of urea's destabilizing effect, and unraveling its influence on tertiary interactions along the folding pathway ( 13 , 36 , 37 ) .In the case of urea denaturation, allows us to compare the energetic impact of urea on the rate constant for docking with its effect on the equilibrium of docking.A value of 0 would indicate that the interaction affected by urea has not formed in the transition state, while a value of 1 would suggest that the interaction affected by urea fully forms in the transition state.A fractional value implies the partial formation of tertiary contacts during the transition state.Our data yield a value of 0.63 for Bsu and value of 0.49 for Tte , indicating that the tertiary interactions perturbed by urea partially form already in the transition state ( Supplementary Table S3 ) .Furthermore, these findings suggest that the docked state is more destabilized by urea than the transition state relative to the undocked state ( 13 ) .
To further investigate the mechanism of the urea-induced unfolding of the two riboswitches, we generated transition occupancy density plots ( TODPs ) .These plots, presented as heat maps ( 34 ) , illustrate the fraction of molecules that undergo a particular transition between defined FRET states at least once during the observation period.With no Mg 2+ but in the presence of preQ 1 , both riboswitches exhibit a negligible contribution from a static mid-FRET population.However, with increasing urea concentration, the occupancy of this population increases ( as observed in the on-diagonal populations in Figure 4 B ) .This suggests that an increasing number of molecules adopt an undocked conformation with minimal dynamic character ( as far as observable at the time resolution of the experiment, 60 ms ) .Comparing the TODPs between Tte and Bsu , we observe that the Tte riboswitch displays a closer proximity between the two interconverting dynamic FRET states, indicating a more compact conformation than for the Bsu riboswitch ( Figure 4 B, left panels ) .Furthermore, at 4 M urea, the Tte riboswitch exhibits increased heterogeneity with multiple underlying FRET values ( Figure 4 B, right panels ) .
In the presence of both 1 mM Mg 2+ and preQ 1 , urea has less impact on the FRET histograms of both riboswitches ( Figure 4 C ) , which is consistent with the stabilizing effect of Mg 2+ .The high-FRET populations of both the Bsu and Tte riboswitches still gradually decrease with increasing urea concentration, although not to the same extent as observed under the condition without Mg 2+ .In particular, the changes in the fractions of high-FRET population are less at 2 M urea when Mg 2+ is present compared to its absence ( Figure 4 C ) .Furthermore, differences in ( un ) folding kinetics were observed between the two riboswitches.Specifically, the undocking rate constant k undock of the Bsu riboswitch remains relatively unchanged upon addition of urea, whereas k undock for the Tte riboswitch significantly increases ( Supplementary Figure S1c, d ) .At the same time, the docking rate constant k dock of the Bsu riboswitch decreases more profoundly than that of Tte ( Supplementary Figure S1c, d ) .Based on transition-state analysis, an unchanged k undock results in a value of 1 for Bsu , indicating that the tertiary interactions perturbed by urea have already been fully established in the transition state upon addition of 1 mM Mg 2+ .TODPs provide additional insights into the differences in transition dynamics.In the presence of Mg 2+ ( with no urea ) , a significant Mg 2+ -induced compaction occurs, resulting in two closely spaced interconverting FRET states ( compare Figure 4 B and D ) .This finding is consistent with the mid-FRET state of the Bsu riboswitch shifting from μ undock ∼0.57 to μ undock ∼0.78 in the presence of Mg 2+ ( compare Figure 4 A and C ) .In contrast, the Tte riboswitch exhibits a high mid-FRET value even without Mg 2+ and preQ 1 , and this value remains largely unchanged upon the addition of Mg 2+ .Overall, Mg 2+ sufficiently stabilizes the pseudoknot in both riboswitches, leading to increased resistance to urea denatu-  ration.This observation is further supported by the relatively small fraction of static mid-FRET state molecules even at 4 M urea ( compare Figure 4 B and D ) .

TMAO promotes riboswitch folding and counteracts urea-induced denaturation
The osmolyte TMAO is known to promote the folded conformation of RNA due to its unfavorable interaction with exposed phosphate groups of the RNA backbone ( 28 ) and is thus expected to counteract urea ( 18 ) .To further compare the folding properties of the Bsu and Tte riboswitches, we studied the effects of TMAO ( in the presence of 100 nM preQ 1 ) on RNA folding, first in the absence of urea.With no Mg 2+ , in case of the Bsu riboswitch, we observed a gradual increase in the mean FRET efficiency of the mid-FRET state from μ undock ∼0.67 in the absence of TMAO to ∼0.83 in the presence of 2 M TMAO, indicating that the osmolyte further compacts the undocked Bsu riboswitch ( Supplementary Figure S2a ) .This effect is smaller ( from 0.75 to 0.87 ) in the presence of both preQ 1 and 1 mM Mg 2+ due to the overall compaction in the Bsu pseudoknot already induced by the divalent metal ion ( Supplementary Figure S2b ) .In contrast, the effects of TMAO on the Tte riboswitch are insignificant ( data not shown ) since the Tte riboswitch is already tightly folded in the presence of only 100 nM preQ 1 , regardless of the presence or absence of Mg 2+ .Furthermore, kinetic analysis of the smFRET time traces reveals an increase in k dock for the Bsu riboswitch with increasing TMAO concentration.This is in contrast to the relatively small effect on k undock of Bsu ( Supplementary Figure S3 ) and either rate constant of the Tte riboswitch ( data not shown ) .
We also investigated the ability of TMAO to counteract the denaturing effects of urea.In the presence of 2 M urea, TMAO titration results in a gradual shift in the μ undock towards higher mid-FRET values for both the Bsu and Tte riboswitches, regardless of the presence or absence of 1 mM Mg 2+ ( Figure 5 A, C ) .Notably, 1 M TMAO is able to counteract 2 M urea, almost restoring the fraction docked to the levels similar to those observed in the absence of urea ( compare Figures 4 and  5 ) .Furthermore, besides inducing overall compaction of both Bsu and Tte riboswitches, we observed an additional effect of TMAO on the Tte riboswitch.It shifts the dynamic transitional population towards a more highly folded static population, as highlighted by the TODPs at high TMAO concentration, which are dominated by the static high-FRET population ( Figure 5 B, D ) .Kinetic analysis shows that, in the absence of Mg 2+ , TMAO increases k dock and decreases k undock , whereas in the presence of Mg 2+ , TMAO only increases k dock without affecting k undock ( Supplementary Figure S4 ) .Together, these observations suggest that TMAO promotes folding of both riboswitches, with the Tte riboswitch exhibiting a more compact overall conformation, as shown by the dominating static high-FRET population ( Figure 5 B, D ) .This highlights the higher tendency of the Tte riboswitch to fold.

Osmolytes-induced changes in folding free energies
From the equilibrium constant of the folding, K eq , calculated from the area under the curve ( AUC ) for the population histograms showing mid-and high-FRET peaks using the equation K eq = ( A UC high FRET / A UC mid FRET ) ( used to be able to consistently capture both the dynamic and static docking behaviors we observe ) , osmolyte-dependent changes in Gibbs free energy of the folding process ( G • fold ) , known as mvalues, were determined.G • fold was then derived using the Gibbs equation ( G • fold = -RTln [K eq ] ) .G • fold appears to be linearly correlated to osmolyte concentration, with the slope of a regression line defined as the m-value.This m-value analysis is commonly performed in osmolyte titration studies where changes in the m-values indicate the sensitivity of folding transitions to the presence of the osmolytes, reflecting the relative amount of solute biopolymer surface exposed to solvent upon unfolding ( 18 ) .
In both the absence and presence of 1 mM Mg 2+ under a 100 nM preQ 1 background, the Bsu riboswitch shows the same degree of destabilization in response to increasing urea concentration, as indicated by the identical m -values of m -Mg 2+ = 0.32 ± 0.04 kcal mol −1 M −1 and m +Mg 2+ = 0.32 ± 0.05 kcal mol −1 M −1 ( Figure 6 a,b ) .In contrast, the Tte riboswitch is less susceptible to urea-induced denaturation, as evidenced by lower m -values ( m-Mg 2+ = 0.22 ± 0.06 kcal mol −1 M −1 and m +Mg 2+ = 0.06 ± 0.03 kcal mol −1 M −1 ) .These findings support our previous observation of a more tightly folded Tte riboswitch structure.Notably, the m -value of Tte in the presence of 1 mM Mg 2+ approaches zero ( m +Mg 2+ = 0.06 ± 0.03 kcal mol −1 M −1 ) , suggesting that Tte folding remains largely unaffected by urea.This observation is consistent with a Mg 2+ -stabilized, preQ 1 -bound structure for Tte , where little additional surface area becomes exposed to solvent upon the addition of urea.
In our TMAO titrations, the m -values for both riboswitches are close to zero, irrespective of the presence of Mg 2+ ( data not shown ) , suggesting minimal additional exposed surface area in the absence of TMAO compared to its presence.However, our previous observation of increasing mean mid-FRET values suggests an overall structural compaction.These two findings imply that structural compaction can occur without much solvent expulsion.Furthermore, we calculated m -values for TMAO's effects on counteracting urea denaturation.Consistent with our previous results, we observed TMAO-induced stabilization of the both pseudoknots, as indicated by negative m -values for both riboswitches under all conditions, except for the Bsu riboswitch in the presence of both Mg 2+ and preQ 1 , which showed an m -value close to zero ( m +Mg 2+ = 0.02 ± 0.15 kcal mol −1 M −1 ) .This observation may be due to that the Bsu riboswitch is already stabilized by Mg 2+ and then further reinforced by preQ 1 binding.Consequently, this structure may exhibit resistance to additional TMAO-induced folding.

Discussion
In this work, using smFRET microscopy coupled with UVmelting studies, we quantitatively compared the distinct kinetic and thermodynamic folding properties of the mesophilic Bsu and thermophilic Tte preQ 1 riboswitch RNAs.We investigated their folding properties in the presence of cognate preQ 1 ligand and the osmolytes urea and TMAO.Thermal denaturation experiments demonstrated that the higher-order structure of the Bsu riboswitch pseudoknot melts more cooperatively with a single broad transition, whereas Tte stems P2 and P1 melt in a stepwise fashion, unveiling a higher thermal stability of the Tte P1 helix in particular ( Figure 2 ) .This observation is consistent with the higher ( 80% ) G-C content of P1 stem in Tte riboswitch compared to the Bsu riboswitch ( 60% ) , revealing one strategy for the evolutionary adaption of a functional RNA from meso-to thermophilic environmental conditions.Our smFRET-based chemical denaturation experiments further uncovered a higher structural stability for the Tte compared to the Bsu riboswitch.Urea denatures both Bsu and Tte riboswitches similarly by decreasing μ undock and μ dock ; however, Tte exhibits a greater resistance to urea-induced denaturation than the Bsu riboswitch ( Figure 4 ) , uncovering beneficial sequence changes in the Tte riboswitch for adaption to a high cellular urea content.Furthermore, we showed that TMAO has only a subtle effect on the conformation of the Bsu riboswitch in the absence of Mg 2+ , which is entirely lost upon addition of Mg 2+ .However, our data also demonstrated that TMAO plays a significant role in counteracting urea-induced structural destabilization by promoting the folded riboswitch conformation through increasing k dock ( Supplementary Figure S4 ) .This indicates how high cellular urea concentrations can be counterbalanced by TMAO production.
Previous studies on urea denaturation have suggested that the addition of urea decreases the global stability of RNA, disrupting both secondary and tertiary structures ( 38 ) .Numerous experimental and theoretical studies have investigated the mechanism underlying urea-induced denaturation of nucleic acids ( 39 ,40 ) .These studies suggest that urea forms favorable interactions with nucleobases, sugars and phosphodiester backbones, which compete with internal RNA interactions ( 18 ) .In bacteria, riboswitches have naturally evolved to express proteins that degrade or transport urea out of the cell in order to minimize its toxic effects ( 41 ) .Driven by the underlying energetic forces, nucleic acids in the presence of urea tend to unfold to expose a larger solvent accessible sur- face area ( S AS A ) and attract more urea molecules ( 42 ) .Consistent with these studies, our results demonstrate that urea destabilizes both the Bsu and Tte riboswitches, shifting the population towards an unfolded conformation.Interestingly, the m -values for Bsu and Tte riboswitches are 0.32 and 0.22 ( Figure 6 ) , respectively, indicating that Bsu is more sensitive to urea and exposes relatively more S AS A upon unfolding compared to Tte .This difference can be attributed to the longer dynamic loop L2 in the Bsu riboswitch, which is largely solvent exposed ( Figure 1 A ) .Furthermore, smFRET allowed us to probe the kinetic origin of urea-induced unfolding, which revealed a decrease in k dock and an increase in k undock , with differential effects on the structurally similar Bsu and Tte riboswitches that are consistent with their mesophilic and thermophilic origins, respectively.
The zwitterionic TMAO also interacts with the nucleobases, sugars, and phosphodiester backbones of nucleic acids, but in an energetically unfavorable fashion.Therefore, exclusion of TMAO from the RNA surface stabilizes a more compact RNA conformation ( 28 ) .Our data show that TMAO compensates for urea-induced denaturation of both riboswitches.Somewhat surprisingly, TMAO only seems to significantly counteract the effect of urea for the Bsu aptamer in the absence of Mg 2+ ( m -value -0.51 versus 0.02 in the absence and presence of Mg 2+ , respectively ) , whereas for the Tte riboswitch, the effects of TMAO and Mg 2+ appear to be additive ( Figure 6 ) .This observation is consistent with TMAO promoting folding primarily through its unfavorable interactions with phosphates, whereas Mg 2+ independently stabilizes folding by bridging two singly negatively charged backbone phosphates.In the case of the Tte riboswitch, additional compaction resulting in less phosphate exposure ( as induced by TMAO ) is nevertheless possible, as evidenced by the negative m-values as a function of TMAO concentration in the presence of 2 M urea.This conclusion is further corroborated by the higher fraction of high-FRET molecules observed for the Tte riboswitch than the Bsu riboswitch in TMAO, even in the presence of urea.Finally, our m-value analysis shows that the Tte riboswitch has a higher tendency to remain folded than the Bsu riboswitch during osmolytic denaturation, consistent with our thermal denaturation studies.Since sensitivity to osmolytes is closely related to the S AS A of a nucleic acid structure, our results suggest that the relative change in S AS A for the Bsu riboswitch is higher than that for the Tte riboswitch.This agrees with the fact that the Bsu riboswitch has a 2-nucleotide larger loop L2 ( Figure 1 A ) that is dynamic in the crystal structure and not as tightly folded as that of the Tte riboswitch ( 10 ,43-45 ) .Such a dynamic and more open conformation of the loop may allow easy penetration of urea into the core of the RNA structure, thereby helping urea denature the Bsu riboswitch.Further support for this notion arises from the observation of a further modified L2 loop sequence in the psychrotolerant Carnobacterium antarcticum preQ 1 riboswitch ( from an organism growing at low temperatures in the Antarctic ) , which was recently found to be sufficiently expansive to accommodate two stacked ligand molecules ( 45 ) .Based on our results, we anticipate that the preQ1 riboswitch from Carnobacterium antarcticum will demonstrate decreased thermal stability and heightened susceptibility to chemical denaturation due to a larger solvent exposure area.It is also known that several marine organisms such as sharks, which feature high cellular urea concentrations, naturally use a 1:2 ratio of TMAO:urea to counteract the denaturation stress of urea.In addition, several investigations have shown that TMAO can rescue proteins from urea-induced denaturation at this and other osmolyte ratios ( 24 ,46 ) .
Our comparative study of two closely related riboswitches from a mesophilic and a thermophilic bacterium builds upon this line of inquiry and provides deep insights into the intricate interplay between RNA conformational dynamics and sequence adaptations in thermophilic organisms during their evolution from mesophilic organisms to thrive in new ecological niches.
first for the high FRET population and the second for the mid-FRET population ) , the area under the curve ( AUC ) for each Gaussian was used to calculate G • dock with the following equation:G • dock = −RT ln K eq = −RT ln ( A UC high F RET /A UC mid F RET ) , with T = 20 • C. The error of G •dock at each osmolyte concentration was estimated again through bootstrapping.Briefly, G • dock was determined for 1000 sets of replicate data, where each set is chosen by sampling M traces from traces in the original dataset with replacement, and where M is equal to the number of traces present in the original dataset.The standard deviation of the resulting 1000 replicates is reported as the error of G • dock at a given osmolyte concentration.The slope of the linear fit ( mvalue ) of G • dock versus osmolyte concentration was determined in OriginLab, where the error of the fitting parameter is taken as the error of the m -value.

Figure 2 .
Figure 2. Probing the thermal stabilities of the str uct urally similar Bsu and Tte riboswitches.Melting curve analysis of the Bsu and Tte riboswitches in the absence and presence of Mg 2+ and preQ 1 ligand ( L ) as indicated.

Figure 3 .
Figure 3. Single-molecule FRET probing of the relative structural stabilities of the Bsu and Tte riboswitches.( A ) Schematic for smFRET-based assessment of osmolyte ( urea / TMAO ) impact on RNA folding through effects on doc king / undoc king kinetics of the two closely related preQ 1 riboswitches.( B ) Top, representative fluorescence intensity vs time trace showing anti-correlated changes in donor ( Cy3 ) and acceptor ( Cy5 ) intensities due to FRET during docking and undocking of helix P2 of the preQ 1 riboswitch.Bottom, changes in the FRET efficiency calculated from the trace, observed here in the absence of Mg 2+ .( C ) smFRET histogram constructed from all FRET values in the dataset observed over the first 100 observation frames, showing two major FRET states fitted with Gaussian distributions that yield mean FRET values as indicated.N , number of molecules.

Figure 4 .
Figure 4. Urea destabilizes the folded conformation of the Bsu and Tte riboswitches.( A ) FRET population distribution histograms of the Bsu ( top panel ) and Tte ( bottom panel ) riboswitches in the absence of Mg 2+ with increasing urea concentration, showing shifts in the mid-FRET and high-FRET populations as quantified in the table.( B ) Transition occupancy density plots ( TODPs ) for the Bsu ( top ) and Tte ( bottom ) riboswitches in the absence of Mg 2+ showing the most common FRET transitions in the absence and presence of 4 M urea.( C , D ) FRET population distribution histograms and TODPs as in panels a and b, respectively, but with 1 mM Mg 2+ added.100 nM preQ 1 was used in all experiments to induce dynamic docking of the molecules.

Figure 5 .
Figure 5. TMAO compacts the undocked state and stabilizes the folded conformation by counteracting urea-induced denaturation.( A ) FRET population distribution histograms of the Bsu ( top panel ) and Tte ( bottom panel ) riboswitches in the absence of Mg 2+ and presence of 2 M urea at increasing TMAO concentrations, showing shifts to a more compact mid-FRET state and a higher population of the high-FRET state.( B ) Transition occupancy density plots ( TODPs ) for the Bsu ( top ) and Tte ( bot tom ) riboswitc hes in the absence of Mg 2+ and presence of 2 M urea, summarizing the FRET transitions in the absence and presence of 1.5 M TMAO.( C , D ) FRET population distribution histograms and TODPs as in panels a and b, respectively, but with 1 mM Mg 2+ added.100nM preQ 1 was present in all experiments to induce dynamic docking of the molecules.

Figure 6 .
Figure 6.m -value analysis reveals distinct folding behaviors of the Bsu and Tte riboswitches.( A, B ) Urea-dependent folding free-energy changes for Bsu ( black ) and Tte ( red ) riboswitches in the absence ( A ) and presence ( B ) of Mg 2+ .( C , D ) TMAO-dependent folding free-energy changes for the Bsu ( black ) and Tte ( red ) riboswitches in the presence of urea and in the absence ( C ) and presence ( D ) of Mg 2+ .100 nM preQ 1 was present in all experiments.Data points in each panel originate from individual titrations from independent experiments performed on different slides.Error bars were derived from the standard deviations of fit coefficients from 1000 bootstrap replicates.The R -squared values for the linear fits of the Bsu and Tte data, respectively, are as f ollo ws: ( A ) 0.96, 0.86; ( B ) 0.95, 0.67; ( C ) 0.85, 0.58; ( D ) 0.01, 0.86.

Table 1 .
The melting temperatures of stems P2 and P1 of the Tte and Bsu riboswitches in the presence and absence of preQ 1 and Mg 2+ as indicated.Under equivalent experimental conditions, the P2 stem of the Bsu riboswitch does not show a separate melting transition