A nascent riboswitch helix orchestrates robust transcriptional regulation through signal integration

Abstract Widespread manganese-sensing transcriptional riboswitches effect the dependable gene regulation needed for bacterial manganese homeostasis in changing environments. Riboswitches – like most structured RNAs – are believed to fold co-transcriptionally, subject to both ligand binding and transcription events; yet how these processes are orchestrated for robust regulation is poorly understood. Through a combination of single molecule and bulk approaches, we discovered how a single Mn 2+ ion and the transcribing RNA polymerase (RNAP), paused immediately downstream by a DNA template sequence, are coordinated by the bridging switch helix P1.1 in the paradigmatic Lactococcus lactis riboswitch. This coordination achieves a heretofore-overlooked semi-docked global conformation of the nascent RNA, P1.1 base pair stabilization, transcription factor NusA ejection, and RNAP pause extension, thereby enforcing transcription readthrough. Our work demonstrates how a central, adaptable RNA helix functions analogous to a molecular fulcrum of a first-class lever system to integrate disparate signals for finely balanced gene expression control.


Introduction
In order to thrive, bacteria must constantly adjust their gene expression to their ever-changing environment.Because of the competition between species, it is crucial for their survival that bacteria adapt quickly to transient nutritional resources as well as external threats such as antibiotics and toxins 1 .
Modulation of gene expression can be achieved in multiple ways.One mechanism bacteria employ is based on riboswitches, structural non-coding RNA elements mostly found in the 5' untranslated regions (5' UTR) of messenger RNA (mRNA) 2 .Bacterial riboswitches exquisitely determine the outcome of gene expression either at the level of transcription elongation or translation initiation 3 .A typical riboswitch contains two interwoven domains: (1) a conserved aptamer region able to recognize a speci c ligand whose binding induces conformational changes and (2) a less conserved expression platform that modulates the expression level of the downstream gene(s) upon signal transduction from the aptamer (Chauvier & Walter 2024, Cell Chem review in press).In the case of transcriptional regulation, ligand binding to the aptamer modulates the formation or disruption of a terminator hairpin in the expression platform that leads to the premature termination of transcription 4 .In addition, it has been shown that structural alterations of the nascent RNA as a function of ligand binding can drive the recruitment of Rho termination factor to the nascent mRNA to promote transcription termination [5][6][7] .
The yybP-ykoY RNA motif, known to sense manganese ions (Mn 2+ ), is a particularly widespread class of bacterial riboswitches that has been found in many human and plant pathogens, and regulates genes associated with Mn 2+ homeostasis through transcription or translation control 8 .It does so by selectively sensing Mn 2+ at sub-millimolar levels over the predominant, millimolar magnesium ion (Mg 2+ ) and any other divalents found in the cell 9,10 .Crystal structures of the riboswitch aptamer domain revealed a pair of helices P1 and P2 stacking on top of each other, as well as the pair of P3 and P4, together forming a four-way junction (4WJ) that sandwiches two binding pockets formed by loops L1 and L3 for selective sensing of each one Mg 2+ and Mn 2+ ion, respectively (Fig. 1a) 9,10 .Single molecule Förster Resonance Energy Transfer (smFRET) studies further revealed that, while the presence of a physiological concentration of Mg 2+ pre-organizes the riboswitch into a dynamically docked global structure, capture of Mn 2+ stabilizes the docked conformation further to up-regulate transcription via antiterminator stabilization 10,11 .Formation of the P1.1 "switch" helix -distal from the Mn 2+ binding site, but proximal to the RNA polymerase (RNAP) -is critical for antitermination since in the absence of cognate ligand its 3' segment partitions toward terminator hairpin formation (Fig. 1a).As with most riboswitches, it is known that the aptamer plays a critical role in initiating the downstream transcription regulation events, however, the biological interplay of co-transcriptional RNA folding, tertiary structure dynamics, switch helix formation, and proximal RNAP is still poorly understood 12 .
For regulatory RNA elements such as riboswitches, co-transcriptional folding events may tightly control the outcome of gene expression, by placing ligand binding and the speed of functional RNA structure adoption in competition with the rate of transcription 5,13 .This competition is expected to lead to kinetic control of gene regulation rather than the thermodynamic control invoked by the global free energy landscape of RNA folding 14,15 , rendering the true biological context challenging to capture.Recent studies of riboswitches in the context of a paused elongation complex (PEC), i.e., in the presence of the DNA template and paused RNAP, have revealed that conformational changes of riboswitches are heavily in uenced by the transcription machinery [16][17][18][19][20][21] , demonstrating the importance of the proximal RNAP for a mechanistic understanding of riboswitch-mediated gene regulation.While determinant, the kinetics of cotranscriptional riboswitch folding are di cult to observe at the short timescale dictated by transcription elongation 12,22 .
Intrinsic RNAP pausing is an off-pathway kinetic step that interrupts the nucleotide addition cycle and can exert an additional layer of transcriptional control 23 .RNAP pause sites have been shown to play important roles in modulating gene expression, including transcription-translation coupling 24 , RNA folding 25,26 , transcription termination 4,27 and transcription factor recruitment 28,29 .RNAP pausing relies on speci c DNA sequences embedded throughout the genome 30 , and can be further promoted by the formation of RNA structures in the vicinity of the RNA exit channel 31 and/or the interaction with speci c transcription factors 32 .N-utilizing substance A protein (NusA) is an essential transcription factor that modulates transcription speed using a multitude of protein and RNA binding interactions.It is required along with other transcription factors for the processive transcription of ribosomal RNA 33,34 and is usurped as a component of the phage lambda protein N-mediated antitermination system 35,36 .NusA can promote transcription termination 27,37,38 , decrease the overall speed of RNAP 19,39 , stimulate hairpin-stabilized pauses [40][41][42] , and participate in the correct folding of the RNA component of RNase P 26,43 .NusA is a multi-domain protein whose N-terminal domain is necessary and su cient for stimulating RNAP pausing by directly contacting the RNAP exit channel 40,41 .This domain is followed by three RNA binding domains S1, KH1 and KH2 40,41,44 and, nally, in Escherichia coli (E.coli), the NusA C-terminal domain with two acidic repeats, AR1 and AR2, that directly interact with the alpha subunit of RNAP 45 .The latter contact induces release of the auto-inhibitory AR2-KH1 interaction, which then allows the S1, KH1 and KH2 domains to also interact with the nascent RNA transcript 46 .Despite this detailed structural understanding and the proximity of NusA and the nascent RNA transcript at the RNAP exit tunnel, to date, there are only few examples for NusA modulating riboswitch-based transcription regulation 13,19,47,48 .Speci cally, it was demonstrated that the time window for ligand sensing is widened in the presence of NusA, therefore allowing for a more e cient gene regulatory response 13,47,48 .However, a recent study has uncovered a dynamic competition between NusA and ligand binding to a uoride-sensing riboswitch, highlighting an active interplay between transcription factor recruitment and co-transcriptional RNA folding within a PEC 19 .
In the present work, we used the paradigmatic, well-studied Mn 2+ -sensing riboswitch from Lactococcus lactis (L.lactis) as a model system to investigate the impact of the transcription machinery on a highly structured nascent RNA.By employing a combination of single molecule and biochemical approaches to a reconstituted PEC, we discovered an intricately balanced mechanism wherein the presence of both RNAP and Mn 2+ assists tertiary structure docking of the riboswitch and thereby actively contributes to the nal gene regulatory outcome through co-transcriptional antiterminator folding.Examining the P1.1 switch helix revealed that it is stabilized by sensing the distal, Mn 2+ -dependent docking as well as the presence of downstream RNAP when a DNA template sequence site-speci cally pauses it in proximity.
We further uncovered a role for transcription factor NusA in favoring P1.1 formation, independent of and redundant, but also competing, with Mn 2+ action.Overall, our results provide a comprehensive mechanism for how a multitude of distinct intracellular signals in uence co-transcriptional folding of a riboswitch by their integration through the central P1.1 switch helix, which acts like the fulcrum of a rstclass lever system to adapt bacterial gene expression dependably to an ever-changing environment.

Ligand-induced stabilization of riboswitch helix P1.1 modulates RNAP pausing e ciency
Transcriptional pausing has been shown to be an inherent parameter that enables co-transcriptional RNA folding and protein recruitment to play an active role in regulating gene expression 23,31 .Therefore, we evaluated how riboswitch folding modulates transcriptional pausing, using single-round transcription assays (Fig. 1b).We identi ed a long-lived intrinsic (elemental) pause site at position G104 in the yybP expression platform, which is four nucleotides downstream of the P1.1 switch helix (Fig. 1b and Supplementary Fig. 1) and closely related to the known bacterial consensus pause sequence in the DNA template 30 .Interestingly, the duration of the G104 pause was signi cantly extended in the presence of saturating Mn 2+ ions (Fig. 1c and Supplementary Table 2), indicating that the ligand-bound state further promotes RNAP pausing at this position.Upon examination of the riboswitch secondary structure relative to the G104 pause position in the RNAP active site, we hypothesized that the formation of the P1.1 switch helix inside the RNAP exit channel stabilizes transcriptional pausing (Supplementary Fig. 1).
To test its involvement in RNAP pausing, we destabilized the P1.1 helix by mutating residues C95 and C96 to G95 and G96, respectively (mutant MP1.1; Fig. 1d and Supplementary Table 2).Consistent with our hypothesis, MP1.1 completely abolishes the impact of Mn 2+ on RNAP pausing e ciency.Importantly, compensatory second-site mutations G5C and G6C (mutant MP1.1cp) not only restore two base pairs at the top of P1.1 but also the Mn 2+ -induced enhancement of RNAP pausing to levels similar to those of the wild-type (Fig. 1d and Supplementary Table 2).
Overall, these data are consistent with a regulatory system in which the ligand-mediated stabilization of RNA helix P1.1 in the RNAP exit channel extends transcriptional pausing.

Ligand-mediated folding of the P1.1 helix depends on the proximity of RNAP
To characterize the co-transcriptional folding of the riboswitch P1.1 helix as a function of ligand binding, we used an in vitro transcription assay that generates a co-transcriptionally folded PEC with a single uorophore at the 5' end for identi cation and subsequent probing at the single molecule level (Fig. 2a) 18 .
To this end, transcription was performed with E. coli RNAP and initiated with a 5' Cy3-labeled dinucleotide (5'-Cy3-ApU) in the presence of only three rNTPs (with rUTP omitted), generating a labeled, active EC halted at position 13 (EC-13), where the rst rUTP would be needed.Upon addition of all four rNTPs, the halted ECs were released synchronously for transcription elongation to continue, with the nascent RNA transcript folding co-transcriptionally prior to probing.To generate a PEC in which the RNAP stops at the G104 pause site, we employed this method on a DNA template harboring an appropriately placed 3' biotin modi cation.Upon incubation with streptavidin prior to the addition of all four rNTPs, the RNAP was stalled at the desired position through the biotin-streptavidin roadblock 5 and can be further immobilized on a polyethylene (PEG)-coated microscope slide for imaging via prism-TIRF microscopy (Fig. 2b).We added a short (8-nucleotide), Cy5-labeled DNA SiM-KARTS (single-molecule kinetic analysis of RNA transient structure) probe against the 5' segment of the P1.1 helix, which allowed for dynamic probing of solvent accessibility of the C2-G8 switching region (Fig. 2b and Supplementary Fig. 1) 19,49,50 .Upon simultaneous excitation of Cy3 and Cy5, single molecules were located on the microscope slide through their Cy3 (green) PEC-104 uorescence signal, whereas the accessibility of the target switching region over time was detected through spikes of the corresponding SiM-KARTS uorescence intensity in the Cy5 (red) channel (Fig. 2c and 2d).The cumulative dwell time distributions of the SiM-KARTS probe in the 5' P1.1 unbound (τ unbound ) and bound (τ bound ) states were tted with double-exponential functions to extract the corresponding binding (k SK−on ) and dissociation (k SK−off ) rate constants, respectively (Fig. 2e   and 2f).
To test the latter notion, we next transcribed PEC-104 in the presence of a saturating concentration (0.5 mM) of Mn 2+ (generally added together with 0.5 mM Mg 2+ for a consistent total concentration of 1 mM divalents) and again probed the 5' P1.1 segment using SiM-KARTS (Fig. 2d).Under these conditions, the probe binding rate constants were 2.93 ± 0.01 x 10 6 M − 1 s − 1 (k SK−on,fast ; fraction 44%) and 0.49 ± 0.01 x 10 6 M − 1 s − 1 (k SK−on,slow ; fraction 56%), whereas the two dissociation rate constants were 0.36 ± 0.01 s − 1 (k SK−off,fast; fraction 85%) and 0.08 ± 0.01 s − 1 (k SK−off,slow ; fraction 15%).Notably, both the faster and slower binding rate constants decreased by ~ 20% compared to the no-ligand condition (Fig. 2e).In addition, the fractional contribution of k SK−on,slow increased (from 41-56%) when co-transcriptionally adding Mn 2+ , at the expense of the k SK−on,fast fraction.Consequently, the weighted average binding rate constant (k SK−on overall ) decreased by ~ 37% (from 2.42 ± 0.17 x 10 6 M − 1 s − 1 to 1.53 ± 0.15 x 10 6 M − 1 s − 1 ) upon addition of Mn 2+ ions (Fig. 2e and Supplementary Table 3).These observations support the notion that ligand binding to the riboswitch in the context of PEC-104 promotes P1.1 helix formation and thus limits SiM-KARTS probe access, albeit not completely.Notably, binding of the SiM-KARTS probe effectively mimics formation of helix P1.1, albeit as an RNA:DNA hybrid (Fig. 2b).The observed slight decrease in overall SiM-KARTS probe dissociation rate constant k SK−off overall (from 0.37 s − 1 to 0.32 s − 1 ; Fig. 2f and Supplementary Table 3) upon addition of Mn 2+ is therefore also consistent with the ligand stabilizing formation of the P1.1 helix.As a control experiment, we performed SiM-KARTS probing of a PEC-104 with our mutant MP1.1 in which nucleotides C95 and C96 were both mutated to G, disrupting the upstream Watson-Crick base pairs of the P1.1 helix (Supplementary Fig. 1).While the overall SiM-KARTS binding rate constant expectedly decreased, there was still a similar decrease (35%) associated with the addition of Mn 2+ (Fig. 2g, Supplementary Fig. 2 and Table S3), suggesting that the ligand in uences the entire P1.1 helix surveyed by the SiM-KARTS probe.
To ask whether the proximal RNAP of PEC-104 participates in the ligand-dependent modulation of the P1.1 helix, we next performed SiM-KARTS probing on a co-transcriptionally folded transcript from which the RNAP was run off by omission of the 3' biotin-streptavidin block.To this end, transcription was performed on an extended DNA template, installing an additional 3' end RNA sequence as "anchor" that was hybridized with a 5' biotinylated capture probe (CP) for subsequent immobilization of the transcript on the microscope slide 19 .Importantly, this released transcript (RT-96) comprised the full yybp riboswitch up to residue C96 to allow for formation of the upper base pairs of the P1.1 helix, mimicking the entire RNA transcript upstream of the RNA-DNA hybrid in the PEC.Strikingly, no effect of Mn 2+ on the RT-96 riboswitch was detected by the SiM-KARTS probe, yielding comparable rate constants in the presence and absence of ligand that match the kinetic parameters for PEC-104 in the absence of ligand, where the P1.1 helix is not (fully) formed (Fig. 2g, Supplementary Fig. 2 and Supplementary Table 3).These observations indicate that the presence of RNAP in PEC-104 actively assists P1.1 helix folding upon addition of Mn 2+ .
Taken together, our single molecule probing reveals that ligand and RNAP, paused at a speci c downstream site, collaborate to nucleate the P1.1 switch helix.Put differently, the riboswitch integrates disparate signals from distal Mn 2+ binding and proximal RNAP pausing for gene regulatory action.
The presence of RNAP promotes the docked conformation of the riboswitch Our biochemical and single molecule probing assays underscore the importance of both RNAP and ligand for riboswitch folding.Generally, the presence of RNAP and DNA template during transcription has been shown to affect folding of structural RNAs, such as riboswitches, which in turn modulates the output of gene expression 5,[16][17][18][19]24 . Theefore, we next evaluated how the proximal RNAP modulates the global docking dynamics when paused at G104 (PEC-104), using smFRET between uorophores attached to P1 and P3 of a hybridized two-strand riboswitch (Fig. 3a and Supplementary Fig. 1), as previously established for the isolated aptamer with stabilized P1.1 10,11 .
At near-physiological concentration of Mg 2+ (1 mM), the riboswitch in PEC-104 sampled two predominant conformational states in which the low-(~ 0.2) and high-FRET (~ 0.7) populations represent the undocked and docked states, respectively (Fig. 3b,d), consistent with previous studies of the isolated aptamer 10,11 .The FRET histogram showed nearly equal populations in the low-and high-FRET regimes (46% and 53%, respectively), indicating that the riboswitch can occupy the docked state signi cantly in the absence of Mn 2+ .Upon addition of 0.5 mM Mn 2+ , the riboswitch adopted a more stably docked conformation (Fig. 3c) with the high-FRET population increasing to ~ 70%, indicating that binding of Mn 2+ shifts the equilibrium toward the docked state (Fig. 3d and Supplementary Table 4), as previously observed absent the PEC 10,11 .
To assess the in uence of RNAP on the global dynamics of (un)docking, we assembled the same FRETlabeled riboswitch and immobilized it to the microscope slide through a biotinylated locked nucleic acid (LNA) capture probe using the same base pairing employed to reconstitute PEC-104 (LNA-104; Fig. 3a).
As illustrated by the presence of the two predominant low-(~ 0.2) and high-FRET (~ 0.7) populations, the riboswitch in the absence of RNAP adopted the same two dynamic conformational states as in PEC-104 (Fig. 3e).Notably, in Mg 2+ only with RNAP removed the docked population was signi cantly less populated, with 42% for LNA-104 compared to 53% under the same conditions for PEC-104 (Fig. 3d and  e), suggesting that the proximal RNAP promotes the formation of the docked state.Similarly, addition of 0.5 mM Mn 2+ led to 65% docked state for LNA-104, but a slightly increased 67% for PEC-104 (Fig. 3d and   e); whereas in 100 mM K + without divalents the riboswitch adopted only 24% docked state for LNA-104, which rose to 35% for PEC-104 (Supplementary Fig. 3 and Supplementary Table 4).These observations demonstrate that RNAP favors docking under a broad set of conditions.
As control, we designed an RNA that went ten bases beyond the G104 pause, and immobilized it through a complementary LNA capture strand (LNA-114) to mimic the structure after full transcription of the riboswitch with RNAP far downstream (Supplementary Fig. 10).In the presence of Mg 2+ only, just 25% of the LNA-114 riboswitch were found in the high-FRET (~ 0.7) docked state, compared to 42% for LNA-104 and 53% for PEC-104 (Supplementary Table 4), suggesting that extension beyond the G104 pause disfavors docking.As expected, with the addition of Mn 2+ the fraction of docked state signi cantly increased to 67% (Supplementary Fig. 10 and Supplementary Table 4), indicating that the full-length riboswitch remains highly Mn 2+ sensitive, even as RNAP has already translocated further downstream.
Taken together, our analysis so far reveals that the presence of RNAP at the G104 pause site promotes the docked riboswitch conformation, particularly in the absence of ligand, due to dynamic sampling of a partially folded P1.1 switch helix.
A semi-docked riboswitch conformation represents the initial nucleation of P1.1 From our previous work it is known that binding of Mn 2+ induces a static docked (SD) conformation in the riboswitch 10 .Hidden Markov modeling (HMM) of smFRET traces obtained from our PEC-14 and LNA-104 con gurations in the absence and presence of Mn 2+ reproduced this behavior.The dynamic docking/undocking of many traces is evidenced as symmetric pairs of off-diagonal contours in the transition occupancy plots (TODP), which represent as a heat map the fraction of single molecules that exhibit speci c initial-to-nal FRET transitions at least once (Fig. 4 and Methods) 51 .In contrast, SD behavior is indicated by a single high-FRET peak on the diagonal, observed most prominently upon addition of Mn 2+ (Fig. 4), consistent with our earlier observations 10 .
Further inspection of the smFRET trajectories revealed that HMM modeling was best performed using As expected from the FRET population histograms, addition of Mn 2+ increased the proportion of the ondiagonal SD population in both PEC-104 (from ~ 20% to ~ 35%; Fig. 4c and 4d) and LNA-104 (from ~ 18% to ~ 46%).As a result, the dynamic population transitioning between the U, D* and D states decreased to ~ 58%, supporting that ligand binding shifts the equilibrium toward a more compact, stably docked conformation.Indeed, the transition rate constant from U to D slightly increased from 2.16 s − 1 in the absence to 2.29 s − 1 in the presence of Mn 2+ , while the transition from D to U became slower (2.41 s − 1 and 1.67 s − 1 in the absence and presence of Mn 2+ , respectively).Binding of Mn 2+ signi cantly shifted the transition towards higher D* and D populations, with a smaller population now sampling the U state (Fig. 4d).Similarly, the U to D* transitions became faster in the presence of Mn 2+ (2.12 s − 1 compared to 1.76 s − 1 with only Mg 2+ ), whereas the transition from D to D* became slower (1.51 s − 1 compared to 1.96 s − 1 with only Mg 2+ ), further supporting that the D* state represents a semi-docked state on route to Mn 2+induced full docking.
In the absence of RNAP (LNA-104), the riboswitch was less dynamic than in PEC-104, preferentially adopting the SU (~ 30%) and SD (~ 46%) populations in the absence and presence of Mn 2+ , respectively (Fig. 4e-h).In addition, k dock was slower and k undock faster than in the absence than the presence of RNAP (Fig. 4f,h), further supporting that proximal RNAP promotes the adoption of the D state.LNA-104 docking was closer to a two-state type transition than PEC-104, with the D* state somewhat less populated.This was particularly true in the presence of Mn 2+ , in which the D* to D transitions increased from 2.44 s − 1 in the presence to 4.16 s − 1 in the absence of RNAP.This observation suggests that the RNAP can stabilize the semi-docked state, possibly through electrostatic and steric interaction with the enzyme as observed in the preQ 1 -sensing riboswitch PEC from B. subtilis 17 .
As a further control, we analyzed the FRET transition rate constants in the extended LNA-114 construct in the absence and presence of Mn 2+ ion (Supplementary Fig. 11).With a now-fully available P1.1 switch helix, the riboswitch behaved similarly to LNA-104, exhibiting a signi cant proportion of SU (42%) and SD (40%) in both the absence and presence of Mn 2+ , respectively.In addition, the D* state was observed to be much less accessed than in both LNA-104 and PEC-104, consistent with the notion that the semidocked conformational state D* may depend on an only partially folded P1.1.Even with three-state HMM ts, the TODP predominantly exhibited two-state transitions between the U and D states (Supplementary Fig. 11).In fact, the overall behavior resembled that of a manganese-sensing riboswitch with a stabilized P1.1 10 , with k undock becoming slower upon ligand binding.Similarly, k undock,D*→U in the presence of Mn 2+ was signi cantly faster for LNA-114 (7.14 s − 1 ) than LNA-104 (4.26 s − 1 ), indicating a less stable D* state when the P1.1 is fully stabilized.This observation also explains why D* was not signi cantly detected in the earlier study featuring the stabilized P1.1 10 .
To further test whether the observed conformational changes are due to an initial nucleation of switching helix P1.1 in PEC-104, we directly surveyed P1.1 folding using smFRET.To access the P1.1 FRET vector, we employed stepwise transcription to uorescently label the 5'end of the transcript with Cy3 and incorporate an azido-modi ed NTP at position 11 (Supplementary Fig. 1), which subsequently was labeled using click-chemistry with DBCO-Cy5 18,52 .To stall the RNAP at position 104 (generating PEC-104), the DNA template was immobilized on streptavidin-coated magnetic beads using 3'desthio-biotin, which allowed for subsequent bead release via a competing biotinylated-DNA oligonucleotide (Supplementary Fig. 12).
At near-physiological concentration of Mg 2+ only (1 mM), helix P1.1 in PEC-104 sampled two predominant conformational states of mid-(~ 0.40) and high-FRET (~ 0.65) values and nearly equal populations (58% and 42%, respectively; Supplementary Fig. 13b).Upon addition of 0.5 mM Mn 2+ , P1.1's high-FRET state became more prominent (Supplementary Fig. 13c), suggesting that it is associated with the tertiary docked state that is similarly stabilized by Mn 2+ , consistent with our previous observations on the other FRET vector (Fig. 3).Notably, the majority of traces were static in one or the other FRET state, populating on-diagonal contours in the TODPs (Supplementary Fig. 13d and e).This nding may indicate either very fast transitions beyond the time resolution of our smFRET acquisition, similar to the short-lived (3 ms) rare ("excited") state of the uoride-sensing riboswitch from B. cereus 53 , or long-lived sterical clashes with the nearby RNAP surface.
Taken together, our single molecule kinetic analyses reveal a previously unobserved semi-docked, prefolded riboswitch conformation involving the initial nucleation of P1.1 and stabilization by both RNAP and ligand binding.
Transcription factor NusA recruitment stabilizes P1.1 but is ejected upon ligand binding Our single molecule and bulk data suggest a mechanism wherein the initial nucleation of P1.1 extends intrinsic RNAP pausing, allowing time for co-transcriptional folding of the riboswitch into its ligandsensing conformation for gene regulation.Transcription factor NusA is known to promote transcriptional pausing in the context of hairpin-stabilized Class I pauses through interaction with both the nascent transcript and the RNAP 40,54,55 .Because folding of P1.1 within the RNAP exit channel could act similarly to a Class I transcriptional pause, we next evaluated RNAP pausing e ciency in the presence of NusA (Fig. 5a).Indeed, in the presence of Mg 2+ only, we observed a ~ 4-fold increase of the G104 half-life upon addition of NusA (210 s in the absence of versus 906 s in the presence of NusA; Fig. 5b,c and Supplementary Table 2).In contrast, in the presence of Mn 2+ NusA no longer has any effect on the G104 pause half-life (859 s in the absence versus 820 s in the presence of NusA).These ndings suggest that the ligand-bound riboswitch with its stabilized P1.1 helix supersedes NusA-mediated enhancement of G104 pausing e ciency, as also seen for the uoride-sensing riboswitch 19 , suggesting redundant action by Mn 2+ and NusA.
Because we identi ed the P1.1 switch helix as the key modulator of the G104 pause in the absence of the transcription factor (Fig. 1d), we next tested the hypothesis that it might also negotiate NusA's ability to affect transcriptional pausing.Indeed, while the MP1.1 mutant completely abolished the effect of Mn 2+ on pausing e ciency (Fig. 1d), the mutant allowed NusA to retain its ability to extend pausing independent of Mn 2+ (Fig. 5c).Complementarily, when the compensatory mutant MP1.1cp was used, the wild-type behavior was restored, i.e., NusA's large effect on pausing in Mg 2+ only was lost upon addition of Mn 2+ (~ 4-fold and only ~ 1.5-fold enhancement of G104 pause half-life by NusA without and with ligand, respectively; Fig. 5c and Supplementary Table 2).These data further support that Mn 2+ and NusA have a redundant function in P1.1 formation.Interestingly, addition of NusA during in vitro transcription of the riboswitch moderately decreased the concentration of Mn 2+ needed to modulate transcription termination and readthrough (Supplementary Fig. 14a and b), suggesting that NusA may not only alter the transcription rate and thus time window for ligand sensing, but also modestly sensitize the riboswitch to its ligand 13,19,47 .In further support, an in vitro termination assay performed in either the absence or presence of NusA revealed a small, but signi cant decrease in transcription readthrough in the presence of both ligand and transcription factor, from 60% in the absence to 43% in the presence of NusA (Supplementary Fig. 14c).
To understand the mechanism by which NusA and Mn 2+ alternatively affect G104 pause behavior, we directly monitored the dynamics of NusA's interaction with PEC-104 through an established single molecule colocalization assay (Fig. 5d) 19 .As for our SiM-KARTS analysis, we generated a cotranscriptionally folded PEC-104 and measured the binding and dissociation kinetics of Cy5-labeled NusA in dependence of Mn 2+ .As observed previously 19 , NusA very transiently bound to PEC-104, as evident from the short spikes of red uorescence signal upon direct excitation of Cy5 (Fig. 5e and 5f).From these spikes, we then calculated the τ bound and τ unbound dwell times of NusA in the absence and presence of ligand (Fig. 5g and 5h).In the presence of Mg 2+ only, NusA bound with two rate constants of 2.10 ± 0.01 x 10 7 M − 1 s − 1 (k NusA−on,fast ; fraction 25%) and 3.06 ± 0.01 x 10 6 M − 1 s − 1 (k NusA−on,slow ; fraction 75%), as well as two dissociation rate constants of 0.75 ± 0.02 s − 1 (k NusA−off,fast ; fraction 91%) and 0.11 ± 0.02 s − 1 (k NusA−off,slow ; fraction 9%; Fig. 5g and 5h and Supplementary Table 5).Similar to the SiM-KARTS probe, these biphasic kinetics suggest that P1.1 exists in (at least) two conformational states that NusA senses.
Interestingly, the binding kinetics resemble those determined for NusA binding to a canonical Class I pause PEC (hisPEC) 19 , further supporting that the G104 pause is stabilized through the formation of the P1.1 switch helix within the RNAP exit channel.
In the presence of Mn 2+ , the cumulative unbound dwell times could only be tted with a singleexponential k NusA−on of 3.35 ± 0.01 x 10 6 M − 1 s − 1 (Fig. 5g), similar to the slower binding rate constant in Mg 2+ only, indicating that Mn 2+ binding leads to a more homogenously folded P1.1 that inhibits the fast recruitment of NusA to PEC-104.The still two dissociation rate constants increased by ~ 2-fold upon addition of ligand, suggesting a competition between Mn 2+ and NusA binding.Overall, these observations demonstrate that ligand binding affects the early step of NusA-mediated enhancement of pausing by preventing initial recruitment of NusA to PEC-104.We were able to link these observations again to P1.1 switch helix formation, since the effect of Mn 2+ on NusA binding was lost for the MP1.1 mutant and largely restored for the compensatory MP1.1cp mutant (Fig. 5i, Supplementary Fig. 15 and Supplementary Table 5).
To ask whether the observed effect of ligand binding on NusA recruitment occurs during cotranscriptional folding, we next surveyed NusA binding during real-time transcription of the riboswitch 19,56 (Supplementary Fig. 16).In Mg 2+ only, and absence of Mn 2+ , NusA remains bound to the active transcription complex for ~ 1.8 s, similarly to our results on the stalled PEC-104 under the same conditions (Supplementary Table 5).In contrast, when Mn 2+ is added co-transcriptionally, a slight, but signi cant decrease of the bound time is observed (~ 1.2 s; Supplementary Fig. 16c), indicating that cotranscriptional folding of the riboswitch in the presence of ligand alters NusA occupancy to the transcription complex in situ.
Collectively, our single molecule and biochemical assays demonstrate that NusA-mediated enhancement of transcriptional pausing is abolished by Mn 2+ -induced nucleation of P1.1, which in turn suppresses NusA recruitment to the PEC.That is, while functionally redundant in extending transcriptional pausing by favoring a stable P1.1, Mn 2+ and NusA do so in mutually exclusive fashion.

NusA stabilizes the transient semi-docked riboswitch conformation
Our single molecule NusA binding assay unveiled that ligand binding to the riboswitch prevents fast NusA recruitment to the PEC through P1.1 stabilization.While NusA binds transiently, as one of the most abundant transcription factors in the bacterial cell (at micromolar concentrations) 57 it is expected to be almost continuously bound to an active elongation complex (EC) in vivo 58 so that it is unlikely that PEC-104 exists in the absence of NusA.To understand the dynamics of the riboswitch in a context closer to in vivo conditions, we next evaluated the global docking dynamics of the NusA-bound PEC-104 using smFRET in the absence and presence of Mn 2+ (Fig. 6a).
Similar to the traces collected for PEC-104 in the absence of Mn 2+ , i.e., in the absence of any divalent ion or with Mg 2+ only, the NusA-saturated PEC-104 shows a high degree of dynamics between the three FRET states U, D* and D (Fig. 6b and Supplementary Fig. 17).In contrast, upon addition of ligand, the smFRET traces become more static, preferentially adopting the D* or D states (Fig. 6c), suggesting that both NusA and ligand stabilize the semi-docked and docked conformation of the riboswitch.Accordingly, in both the absence and presence of Mn 2+ with NusA bound the cumulative FRET histograms could be best tted with three independent Gaussian peaks, corresponding to the three conformational states (U, D* and D; Fig. 6d and f).Moreover, upon Mn 2+ addition, the equilibrium shifted from the U toward the D* state, rather than adopting the D state, further supporting that NusA particularly favors this intermediate semidocked state.
Kinetic analysis of the transitions between the different conformational states further con rmed that the riboswitch was mostly dynamic in the absence of ligand.However, while the proportion of static molecules remained the same (29% in the absence versus 28% in the presence of NusA) with only Mg 2+ , the transitions from U to either the D* or D states became much faster in the NusA-bound PEC compared to PEC-104 in the absence of the transcription factor (compare Fig. 6e and Fig. 4b and Supplementary Figs. 5 and 19).Indeed, both k dock,U→D and k dock,U→D* signi cantly increased in the presence of saturating transcription factor, suggesting that NusA promotes fast dynamics into both states (Fig. 6e).Further, the transitions between the D* and D states were observed to be the most probable (64%) with a faster regime toward the D state (5.63 s − 1 ; fraction 34% in the presence of NusA versus 3.03 s − 1 ; fraction 17% in the absence of NusA).
Upon addition of Mn 2+ , the transition kinetics between the conformational states revealed that a 25% static population was found in SD* and 30% in SD (Fig. 6g), suggesting that NusA assists Mn 2+ -induced stable docking.Most of the dynamic transitions (~ 64%) were observed between the D* and D states, with faster transitions toward the D state compared to PEC-104 without NusA-bound (4.66 s − 1 ; fraction 36% in the presence of NusA versus 2.44 s − 1 ; fraction 18% in the absence of NusA; Fig. 6g and Supplementary Fig. 20).Conversely, the transitions from the D to the D* state were signi cantly slower in the presence of both Mn 2+ and NusA (1.03 s − 1 ; fraction 28% in the presence of NusA versus 1.51 s − 1 ; fraction 16% in the absence of the transcription factor), supporting that NusA promotes full docking of the riboswitch through stabilization of the P1.1 switch helix.In further support, SiM-KARTS probing of P1.1 performed on NusA-bound PEC-104 revealed that this region is less accessible in the presence of the transcription factor (Supplementary Fig. 21 and Supplementary Table 3), con rming its impact factor on the local stabilization of P1.1.
NusA preferentially recognizes the single-stranded P1.1 in the ligand-free conformation Our single molecule binding assays and structural probing so far unveiled a previously unappreciated effect of NusA transcription factor on nascent RNA structural dynamics and vice versa (Figs. 5 and 6).While the P1.1 switching helix is central for the overall gene regulation mediated by ligand binding, from transcriptional pausing to transcription termination, it is still unclear which conformational state is preferred for the initial recruitment of NusA.To probe the mechanistic details of NusA-mediated regulation of gene expression as a function of RNA structure, we next evaluated the kinetics of NusA binding to an "open" P1.1 helix using a 3-color microscopy experiment.Here, the open (single-stranded) state of P1.1 was surveyed through Cy3-labeled SiM-KARTS probe binding (green), while simultaneously observing the binding of Cy5-labeled NusA (pink) to PEC-104 (blue, Fig. 7a and b).
In the presence of only Mg 2+ , NusA colocalized with the undocked P1.1 helix for ~ 1.4 s, a value very similar to the overall bound time of NusA to PEC-104 under the same conditions (Supplementary Table 5), supporting the notion that NusA captures the open P1.1 during its recruitment.Strikingly, upon addition of Mn 2+ , this colocalization dwell time signi cantly decreased by ~ 2.5 fold (Fig. 7c), suggesting that folding of the P1.1 helix in the presence of ligand and NusA occupancy are mutually exclusive.Control experiments surveying colocalization dwell times in the MP1 mutant of PEC-104 further corroborated that the effects on NusA association kinetics observed upon ligand binding require the ability to form the top base pairs of P1.1 (Fig. 7d).
Taken together, we discovered a previously unappreciated effect of NusA transcription factor on RNA structural dynamics, wherein binding of NusA to PEC-104 stabilizes an intermediate, semi-docked state of the riboswitch to assist downstream folding in the ligand-bound conformation.All of these effects involve the P1.1 switch helix as a focal point on which Mn 2+ , NusA and RNAP all act; in return, P1.1 integrates these disparate signals as it modulates gene expression.

Discussion
In this work, we have mechanistically dissected the co-transcriptional folding of the L. lactis Mn 2+ riboswitch, representative of the widespread yybp family, using a combination of transcription and single molecule probing assays.First, we investigated the pausing behavior of RNAP and found it to be signi cantly extended by co-transcriptional binding of either the cognate riboswitch ligand Mn 2+ (Fig. 1) or transcription factor NusA (Fig. 5).Second, using single molecule secondary structure probing by SiM-KARTS we discovered that Mn 2+ binding together with the proximal RNAP signi cantly favor formation of the central switch helix P1.1 (Fig. 2).Third, using smFRET we resolved three global riboswitch conformations: undocked (U), semi-docked (D*) and docked (D), where Mn 2+ and proximal RNAP signi cantly stabilize the D* and D states over U (Figs. 3 and 4), whereas NusA stabilizes mostly D* (Fig. 6).Fourth, NusA primarily recognizes the unfolded P1.1 (Fig. 7), leading to its rejection from the paused elongation complex once Mn 2+ binds to stabilize both docking and P1.1 formation (Figs. 4, 5 and   S13).
Taken together, our observations lead to the model in Fig. 8, wherein the signals of ligand binding and the ensuing tertiary structure docking of the aptamer domain, pausing by RNAP on a speci c downstream DNA template sequence, and NusA transcription factor binding to the paused elongation complex are integrated by the central P1.1 switch helix to affect gene regulation.In this way, transcription-associated processes help ne-tune in real-time a local-to-global-to-local hierarchy of conformational dynamics of the nascent riboswitch during adaptive transcription.Within this molecular machine, the top two base pairs of the P1.1 helix fold in the RNAP exit channel and act as a central, nanoscale fulcrum, analogous to a rst-class lever system, to nely balance all incoming signals for adaptive partitioning into either termination (without Mn 2+ ) or antitermination (with Mn 2+ ; Fig. 8).This molecular fulcrum mechanism empowers a bacterium to rapidly respond to an environmental threat by producing an exporter of excess Mn 2 + 12 .Our mechanistic probing thereby signi cantly advances our understanding of the complex, tightly interwoven regulatory interaction network inside the bacterial cell that is all sensed and processed by the top base pairs of the P1.1 switch helix.RNAP pausing has been known for its potential to facilitate co-transcriptional riboswitch folding, allowing time for the transcript to adopt a speci c and functional conformation 5,17,18,59 , and/or to bind its cognate ligand 13,47 .Transcriptional pausing classically has been understood to be stabilized by two main mechanisms: hairpin-stabilized (class I) and backtracked (class II) 32 .Class I pauses are further stabilized by transcription factor NusA upon interaction with the nascent RNA paused hairpin and the β-ap domain of the RNAP, resulting in the stabilization of the swiveled conformation of the RNAP 40,60 .Class II pauses instead are due to a reverse motion of the RNA relative to the enzyme (backtracking) and lead to the disengagement of the RNA 3' end from the RNAP catalytic site.This class of pause is sensitive to GreB and NusG, which promote RNA cleavage and suppress backtracking, respectively, for subsequent transcription reactivation 61,62 .More recently, a third class of transcriptional pausing (class III) has been identi ed in the preQ 1 -sensing riboswitch from B. subtilis, where folding of an RNA pseudoknot structure inside the RNA exit channel modulates RNAP pausing e ciency distinctly in the absence and presence of ligand 17,21 .Our results here demonstrate that the Mn 2+ riboswitch G104 pause shares some similarities with the canonical class I pause, in that the nucleation of the top two base pairs of P1.1 stabilizes RNAP pausing.Notably, these base pairs reside at the edge of the RNA-DNA hybrid, within the RNA exit channel, in a con guration similar to the canonical class I his pause hairpin 63 .We also found that the G104 pause is enhanced in the presence of Mn 2+ ligand (Fig. 1), in parallel to the ligand-mediated stabilization of the P1.1 helix, analogous to the pausing mechanism found in the uoride-sensing riboswitch 19 .In contrast to the class I pause, however, NusA is ejected from PEC-104 upon ligand binding and P1.1 stabilization (Fig. 5), perhaps because of a sterical clash with the docked remainder of the aptamer (Fig. 8).The rst two G:C base pairs nucleate formation of the complete P1.1 switch helix and suppress the downstream strand invasion needed to form the overlapping terminator hairpin, reminiscent of the strand invasion mechanisms identi ed for other transcriptional riboswitches 64-66 (Fig. 8).These observations support a key role for transcription pausing exactly at G104 in setting the stage for adaptable P1.1 folding to regulate downstream gene expression.
The structure of the L. lactis Mn 2+ riboswitch has been solved previously by X-ray crystallography 9 .In addition, the structure and dynamics of the closely related Xanthomonas oryzae Mn 2+ riboswitch were determined, highlighting the importance of adjacent Mg 2+ and Mn 2+ binding sites for tertiary structure docking of this elemental ion-sensing riboswitch in isolated form 10 .We found that the proximity of RNAP during co-transcriptional RNA folding, known to affect a range of RNA structure-based processes [67][68][69] , is also a critical contributor to the ability of the nascent Mn 2+ riboswitch to regulate the outcome of gene expression, by stabilizing both global tertiary structure docking and local secondary structure (P1.1 helix) formation.Bacterial RNAP paused on its DNA template previously also has been shown to in uence the adoption of transient transcriptional intermediates, allowing for more controlled sequential folding toward the regulatory structure 5,18,70 .Similarly, using smFRET we were able to determine that the intermediary, only semi-docked riboswitch state D* becomes prevalent in the PEC at the G104 pause site, perhaps indicating a direct steric clash with the proximal RNAP surface (Fig. 8).These ndings are reminiscent of the folding pathway of the uoride-sensing riboswitch in a similar paused complex 16 , supporting a broad in uence of RNAP on riboswitch folding.Such metastable conformational states additionally have been identi ed in other riboswitches, even in isolated form without RNAP 59,71,72 , suggesting that pre-organization is a widespread strategy deployed by riboswitches for e cient ligand binding and folding in service of decisive regulation of gene expression.Moreover, we found that the short-lived D* state is further stabilized by NusA binding to PEC-104, especially in the presence of ligand (Fig. 6).From the broader perspective of RNA folding in vivo, this observation highlights a versatile role of transcription factor NusA, proposed to assist the nucleation of RNA duplexes within the RNA exit channel 40 , beyond the stabilization of imperfect intrinsic terminator hairpins 38,55 .Speci cally, for the Mn 2+ riboswitch it performs a regulatory function upstream of the transcriptional decision point, similar to its role for the elemental uoride ion-sensing riboswitch 19 .In addition, the D* state is undetectable once the entire P1.1 sequence is available upon release from or downstream transcription by the RNAP (Figs. 3 and 4).Taken together, these observations support a sensitive transcription regulation checkpoint at the G104 pause where ligand binding and the transcription machinery act in concert through fulcrum helix P1.1 to facilitate a nely balanced partitioning of the downstream expression platform into transcription termination or readthrough, depending on the input from multiple, integrated signals (Fig. 8).
The mechanistic single molecule approaches developed here will undoubtedly help dissect many more intricate dynamic interplays between RNA folding and transcription, anticipated to govern the function of numerous non-coding RNAs across bacteria (Chauvier & Walter 2024, Cell Chem Review in press), with the potential to be leveraged for the design of novel anti-bacterial drugs 73 .

Limitations of the study
This study faced two main limitations.First, due to the absence of an atomic structure of the paused elongation complex studied here, we lack a detailed picture of the diverse interactions within the RNAP's exit channel that involve both RNA transcript and NusA transcription factor.Structural determination of such complexes will be needed to pinpoint the interfaces involved in such an intriguing mechanism of pausing and unpausing.Second, while docked and undocked P1.1 could be observed using a FRET vector directly surveying P1.1 structure, only few transitions were observed between the two FRET states.
We therefore could not directly link the dynamics of the P1.1 switch helix with those of the global conformational states U, D* and D. We hypothesize that the P1.1-associatedrate constants might be beyond the time resolution of the camera used to collect the smFRET data.

DNA templates
A 249-nucleotides DNA template including the yybp riboswitch from L. lactis under the control of the T7A1 promoter was cloned into pUC19 plasmid between EcoRI and BamHI restriction sites.Transcription templates for in vitro transcription were generated by PCR using the "T7A1-PCR" forward oligonucleotide and the according reverse oligonucleotides.For stepwise transcription, the DNA template was generated by PCR using the "T7A1-Yybp-Start_AC" forward oligonucleotide and the "Yybp-EC104-DesthioBiotin" reverse oligonucleotide.For mutant DNA templates, the mutation has been inserted in two PCR steps using overlapping oligonucleotides containing the corresponding mutation.For the evaluation of the G104 pause mutants to the termination of transcription, because the 3' segment of P1.1 is also part of the terminator stem, MP1.1 and MP1.1cp shows only transcription readthrough regardless of the presence of Mn 2+ .Therefore, for the transcriptions presented in Supplementary Fig. 14c, the DNA templates also contain a mutation in the terminator stem allowing base pairing in the mutant contexts.
Reaction aliquots were denatured before loading 5 µL of each onto a denaturing 8 M urea, 6% polyacrylamide sequencing gel.The gel was dried and exposed to a phosphor screen (typically overnight), which was then scanned on a Typhoon Phosphor Imager (GE Healthcare).

Transcription data analysis
To determine the T 50 for regulation of termination by Mn 2+ , the relative intensity of the full-length product was divided by the total amount of RNA transcript (full-length + terminated product) for each Mn 2+ concentration.The resulted percentages of readthrough were plotted against the Mn 2+ concentration using equation: Y = Y 0 + ((M1×X)/(T 50 + X)), where Y 0 is the value of Y at X min (no ligand condition) and The half-life of transcriptional pausing was determined by calculating the fraction of each RNA pause species compared with the total amount of RNA for each time point, which was analyzed with pseudorst-order kinetics to extract the half-life 74 .For each determination, we have subtracted the background signal.Error bars in transcription quanti cation represent the standard deviation of the mean from independent replicates.

NusA expression and puri cation
Wild-type and single-cysteine NusA were expressed from plasmids pNG5 and pKH3 respectively in BLR (DE3) cells.The cells were grown in LB media supplemented with 50 µg/mL kanamycin, induced with 1 mM IPTG when OD 600 reached ~ 0.7 and harvested 3 hours post-induction.His-Tagged proteins were puri ed using nickel a nity chromatography as described previously 41 .The fractions containing NusA protein were con rmed by SDS-PAGE and dialyzed into low salt dialysis buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5 mM TCEP, 0.1 mM Na 2 EDTA) and subsequently puri ed using ion-exchange chromatography (Mono Q) using a NaCl gradient from 0.2 M to 2 M. In the case of the single-cysteine mutant, the ion-exchange chromatography was performed after the labeling step (see below).After puri cation, the protein was found to be puri ed to > 95% homogeneity and mixed in 1:1 ratio with storage buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5 mM TCEP, 0.1 mM Na 2 EDTA, 50% glycerol) and ash frozen in liquid nitrogen for storage at -80°C.

NusA labeling
NusA was labeled with Cy5-maleimide dye (PA25031, GE Healthcare) using an 8:1 molar ratio of dye to protein in a total volume of 200 µL (~ 20 µM protein) of labeling buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5 mM TCEP, 0.1 mM Na 2 EDTA).After ~ 4.5 hours incubation at 4°C, the reaction was quenched upon addition of excess 2-mercaptoethanol.NusA-Cy5 and the free dye were separated by ion exchange chromatography.The labeling reaction was loaded on Mono Q column and washed with excess amount (over 20 column volume) of ion exchange buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5 mM TCEP, 0.1 mM Na2EDTA) to remove the free dye.Then NusA-Cy5 was eluted from the column using a NaCl gradient from 0.2 M to 2 M. The fractions containing NusA protein were con rmed by SDS-PAGE, the puri ed labeled protein was mixed in 1:1 ratio with storage buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5 mM TCEP, 0.1 mM Na 2 EDTA, 50% glycerol) and ash frozen for storage at -80°C.Labeling stoichiometry was determined as ~ 0.6 for Cy5/NusA using the absorbance measurements at A 280 and A 650 .

Preparation of uorescently labeled nascent transcripts
In vitro transcription reactions were performed in two steps to allow the speci c incorporation of Cy3 at the 5-end of the RNA sequence.Transcription reactions were performed in the same transcription buffer described above.Transcription reactions were initiated by adding 100 µM 5'-Cy3-ApU dinucleotide or 5'-ALEXA488-ApU for 3-color experiments and 25 µM ATP/CTP/GTP nucleotides at 37°C for 10 min, thus yielding a uorescent halted complex.The sample was next passed through a G50 column to remove any free nucleotides, and the transcription was resumed upon addition of all four rNTPs at the indicated concentration and heparin (450 µg/mL) to prevent the re-initiation of transcription.The resulted nascent transcript was hybridized to the 5' biotinylated CP (Anchor Bio oligonucleotide) complementary to the 3' end capture sequence, allowing immobilization of the complex to the microscope slide.The CP was mixed to a ratio of 10:1 with RNA transcript and added for 5 min before owing the whole complex to the microscope slide.In the case of PEC transcription, the DNA templates contain a biotin at the 5' end of the template DNA strand.Streptavidin was mixed to a ratio of 5:1 with the DNA template for 5 min prior to start the transcription reaction.Mn 2+ (when present) was always added co-transcriptionally at a nal concentration of 0.5 mM, and the same concentration of ligand was added into the corresponding buffer during subsequent dilutions.

Preparation of uorescently labeled PEC for smFRET
Two uorescently labeled strands for the Mn 2+ sensing riboswitch from L. lactis were purchased from GE Darmacon.The two RNA strands were combined at a nal concentration of 1 µM with either capture probe (LNA) or template DNA (tDNA) and non-template DNA (ntDNA) in nucleic acid scaffold buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl) at 1:1:1:1 ratio (RNA1:RNA2:tDNA:ntDNA).The solution was heated at 90°C for 2 minutes followed by incubation at 37°C for 10 minutes and then at room temperature for 10 minutes before storing at 4°C.This annealed complex was stored at -20°C for future experiments.
For experiments in the absence of RNAP, the annealing mixture was diluted at a concentration of 15-50 pM RNA and owed onto a slide that had previously been passivated with a 10:1 ratio of PEG to biotinylated PEG, then incubated for 10-15 min with a 0.2 mg/mL solution of streptavidin.
For experiments performed in the presence of RNAP, the annealed nucleic acid scaffold was diluted to 50 nM RNA in assembly buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl, 1 mM MgCl 2 ), and E. coli RNAP was added to 250 nM.The mixture was incubated at 37°C for 15 minutes and stored at 4°C for the duration of single-molecule experiments.50 pM of the reconstituted PEC was added to the slide for single-molecule imaging.

Single molecule experiments
All single-molecule uorescence microscopy experiments were performed using a prism-based TIRF microscope based on an Olympus IX-71 frame equipped with a 60X ~ water-immersion objective (Olympus UPlanApo, 1.2 NA) or using the Oxford Nanoimager (ONI) microscope in TIRF (total internal re ection uorescence) mode for 3-color experiments.All SiM-KARTS and NusA colocalization movies were collected at 100 ms time resolution and smFRET movies were collected at 60 ms time resolution using an intensi ed charge-coupled device camera (Hamamatsu C13440-20CU scienti c complementary metal-oxide semiconductor camera).PEG-passivated quartz slides with a micro uidic channel containing inlet and outlet ports for buffer exchange were assembled as described in previous works 50,75 .The surface of the micro uidic channel was coated with streptavidin (0.2 mg/mL) for 10 to 15 min prior to owing the nascent transcripts hybridized to the CP or PEC for smFRET.In the case of nascent transcripts analysis, the uorescent PECs were directly injected into the channel surface using the biotin-streptavidin roadblock for immobilization.
For SiM-KARTS experiments, the nascent RNA transcripts were diluted in SiM-KARTS buffer (80 mM HEPES-KOH, pH 7.5, 300mM KCl).For NusA colocalization assays, the molecules were diluted in imaging buffer (40 mM Tris-HCl, pH 8.0, 330 mM KCl, 0.1 mM EDTA, 0.1 mM dithiothreitol, 1 mg/mL bovine serum albumin).NusA-Cy5 (2 nM) or the SiM-KARTS probe (12.5 nM) was injected into the channel in the corresponding imaging buffer along with an enzymatic oxygen scavenging system consisting of 44 mM glucose, 165 U/mL glucose oxidase from Aspergillus niger, 2,170 U/mL catalase from Corynebacterium glutamicum, and 5 mM Trolox to extend the lifetime of the uorophores and to prevent photo-blinking of the dyes.The raw movies were collected for 15 min with direct green (532 nm) and red (638 nm) laser excitation for SiM-KARTS and NusA colocalization assays and 3 min with direct green (532 nm) for smFRET.

Data acquisition and analysis for SiM-KARTS and NusA colocalization assays
Locations molecules and uorophore intensity traces for each molecule were extracted from raw movie les using custom-built MATLAB codes.Traces were manually selected for further analysis using the following criteria: single-step photobleaching of Cy3, ≥ 2 spikes of Cy5 uorescence of more than 2fold the background intensity.Traces showing binding events were idealized using a two-states (bound and unbound) model using a segmental k-means algorithm in QuB 51 .From the idealized traces, dwell times of NusA or the SiM-KARTS probe in the bound (τ bound ) and the unbound (τ unbound ) states were obtained.Cumulative plots of bound and unbound dwell-time distributions were plotted and tted in Origin lab with single-exponential or double-exponential functions to obtain the lifetimes in the bound and unbound states.The dissociation rate constants (k off ) were calculated as the inverse of the τ bound , whereas the association rate constants (k on ) were calculated by dividing the inverse of the τ unbound by the concentration of SiM-KARTS probe used during the data collection.

Data acquisition and analysis for smFRET
Locations of molecules and uorophore intensity traces for each molecule were extracted from raw movie les using custom-built MATLAB codes.Single-molecule traces were then visualized using custom Matlab code and only those with a minimum combined intensity (Cy3 + Cy5 intensity) of 300 A.U., showing single-step photobleaching of the dyes, a signal-to-noise ratio of > 3, and longer than 6 s were selected for further analysis.Selected traces were then background-subtracted to correct for crosstalk and (minimal) bleed-through.We calculated the FRET ratio as I A /(I A +I D ), where I A and I D are the background-corrected intensities of the acceptor (Cy5) and donor (Cy3), respectively.FRET histograms were made using the rst 100 frames of all traces in each condition and t with a sum of Gaussians using OriginPro 8.5.For kinetic analysis, traces were idealized with a three-state model corresponding to Undocked (low-FRET), Semi-Docked (mid-FRET) and Docked (high-FRET) states using the segmental k-means algorithm in QuB software as previously described 51 .Cumulative dwell-time histograms were plotted from all extracted dwell times and t with single-or double-exponential functions using OriginPro 8.5 to obtain the lifetimes in the undocked (τ undock ) and docked (τ dock ) states.Rate constants of docking and undocking were then calculated as k dock = 1/τ undock and k undock = 1/τ dock .For the double-exponential ts, kinetics were calculated similarly using both the short and long dwell lifetimes to obtain the fast and slow rate constants, respectively.The idealized smFRET traces were used for creating transition occupancy density plots (TODPs), which show the fraction of traces/molecules that exhibit a given type of transition at least once.In TODPs, dynamic traces showing a FRET transition (regardless of the number of transitions in that trace) and static traces (with no transitions over the entire trace) are weighted equally, avoiding over-representation of the traces with fast transitions.

Declarations Figures
three FRET states (Undocked = U (~ 0.2), Semi-Docked = D* (~ 0.45), Docked = D (~ 0.75)).In the presence of Mg 2+ only, ~ 70% of all riboswitch molecules in PEC-104 sampled these three FRET states dynamically over minor stable undocked (SU, ~ 9%) and SD populations (~ 20%; Fig.4a and 4b).Transitions from the U to the D state (k dock ) and from D to U (k undock ) occurred at comparable rate constants of ~ 2.2 s − 1 and ~ 2.4 s − 1 , implying facile reversibility of docking.In the absence of RNAP in LNA-104, the rate constants shifted to ~ 1.1 s − 1 (k dock ) and ~ 5.5 s − 1 (k undock ; Fig.4e and 4f), consistent with our previous conclusion that RNAP promotes docking.Likewise, transitions from U or D to D* and from D* to U or D were more prominent in PEC-104 than in LNA-104, suggesting that the semi-docked D* is also favored by the RNAP.A relatively slow k dock,U→D* transition of ~ 1.8 s − 1 and a faster k undock,D*→U of ~ 2.9 s − 1 in PEC-104, combined with an even faster k dock,D*→D of ~ 3.03 s − 1 , suggest that the D* state is a frequent, albeit nonobligatory, intermediate between the undocked and undocked states.These ndings show that, in the presence of a near-physiological Mg 2+ concentration, the riboswitch within the PEC can sample the previously unobserved semi-docked state D*, in addition to the previously reported U and D states.

Folding
of P1.1 helix upon Mn 2+ binding transcriptional pausing (a) The Mn 2+ -sensing riboswitch from L. lactis regulates gene expression at the transcriptional level.Stabilization of the P1.1 helix upon Mn 2+ binding (Mn) prevents the formation of an intrinsic terminator hairpin, leading of the transcription of the downstream gene.Position of the G104 pause site is indicated as an orange star on the secondary RNA structure.(b)Representative denaturing gel showing the RNAP pauses during the transcription of the Mn 2+ riboswitch.Position of the pause (G104), termination (T) and full-length (FL) products are indicated on the right.Experiments were performed using 25 µM rNTPs in the absence (-Mn 2+ ) and presence (+Mn 2+ ) of 0.5 mM Mn 2+ .The chase lanes (Ch) were taken at the end of the time course after an additional incubation with 500 µM rNTPs for 5 min.(c)Fraction of complexes at the G104 pause as a function of the transcription time in the absence (blue) and presence (orange) of 0.5 mM Mn 2+ .The reported errors are the SD of the mean from independent replicates.(d)Quanti cation of the G104 pause half-life in the WT, MP1.1 and MP1.1cp riboswitch variants.Experiments were performed using 25 µM rNTPs in the absence (blue) and presence (orange) of 0.5 mM Mn 2+ .Error bars are the SD of the mean from independent replicates (***p < 0.01, **p < 0.05, *p < 0.1).The predicted conformation of the P1.1 is represented below each riboswitch variant dataset.

Figure 3 Investigation
Figure 3

Figure 4 Effect
Figure 4

Figure 5 Effect
Figure 5