A nascent riboswitch helix orchestrates robust transcriptional regulation through signal integration

Ligand-induced stabilization of riboswitch helix P1.1 modulates RNAP pausing efficiency

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^27,35. Therefore, we evaluated how riboswitch folding modulates transcriptional pausing, using single-round transcription assays (Fig. 1b). We identified 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 template³⁴. Interestingly, the duration of the G104 pause was significantly extended in the presence of saturating Mn²⁺ ions (Fig. 1c and Supplementary Table 2), indicating that the ligand-bound state further promotes RNAP pausing at this position. Control experiments performed with a mutant abolishing Mn²⁺ binding to the riboswitch (A41U)¹³ showed no such difference, supporting that promotion of RNAP pausing is the result of riboswitch folding in the ligand-bound state, rather than a direct effect on the enzyme (Supplementary Table 2).

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 Mn²⁺ on RNAP pausing efficiency. 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 Mn²⁺-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 fluorophore at the 5’ end for identification and subsequent probing at the single-molecule level (Fig. 2a, b)²². 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)^23,53,54. Upon simultaneous excitation of Cy3 and Cy5, single-molecules were located on the microscope slide through their Cy3 (green) PEC-104 fluorescence signal, whereas the accessibility of the target switching region over time was detected through spikes of the corresponding SiM-KARTS fluorescence intensity in the Cy5 (red) channel (Fig. 2c, d). The cumulative dwell time distributions of the SiM-KARTS probe in the 5’ P1.1 unbound (τ_unbound) and bound (τ_bound) states were fitted with double-exponential functions to extract the corresponding binding (k_SK-on) and dissociation (k_SK-off) rate constants, respectively (Fig. 2e, f).

In the presence of near-physiological 1 mM Mg²⁺ only, the SiM-KARTS probe bound to its 5’ P1.1 target with two rate constants of 3.67 ± 0.02 × 10⁶ M^−1 s^−1 (k_SK-on,fast; fraction 59%) and 0.63 ± 0.01 × 10⁶ M^−1 s^−1 (k_SK-on,slow; fraction 41%), as well as dissociated with two rate constants of 0.55 ± 0.01 s^−1 (k_SK-off,fast; fraction 61%) and 0.08 ± 0.01 s^−1 (k_SK-off,slow; fraction 39%). This observation is consistent with the notion that the SiM-KARTS probe senses (at least) two alternative structures of the riboswitch P1.1 helix that could arise on the same RNA molecule. Indeed, the targeted C2-G8 region is expected to either form the P1.1 helix or adopt a single-stranded conformation, which the absence of Mn²⁺ ligand should favor^12,13.

To test the latter notion, we next transcribed PEC-104 in the presence of a saturating concentration (0.5 mM) of Mn²⁺ (generally added together with 0.5 mM Mg²⁺ 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 × 10^6 M^−1 s^−1 (k_SK-on,fast; fraction 44%) and 0.49 ± 0.01 × 10⁶ 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% to 56%) when co-transcriptionally adding Mn²⁺, 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 × 10⁶ M^−1 s^−1 to 1.53 ± 0.15 × 10⁶ M^−1 s^−1) upon addition of Mn²⁺ 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 Mn²⁺ 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 Mn²⁺ (Fig. 2g, Supplementary Fig. 2 and Supplementary Table 3), suggesting that the ligand influences 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 slide²³. 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 Mn²⁺ 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 Mn²⁺.

Taken together, our single-molecule probing reveals that ligand and RNAP, paused at a specific downstream site, collaborate to nucleate the P1.1 switch helix. Put differently, the riboswitch integrates disparate signals from distal Mn²⁺ 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. However, SiM-KARTS only reveals structural rearrangements that occur locally (P1.1 stem) and does not report the intramolecular dynamics for a single RNA molecule. Generally, the presence of RNAP and DNA template during transcription has been shown to affect global folding of structural RNAs, such as riboswitches, which in turn modulates the output of gene expression^{7,20,21,22,23,28}. Therefore, we next evaluated how the proximal RNAP modulates the global docking dynamics when paused at G104 (PEC-104), using smFRET between fluorophores 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 P1.1^13,15.

a smFRET experimental setup. Paused Elongation Complex (PEC) immobilization utilizes biotinylated E. coli RNAP. Locked Nucleic Acid (LNA) immobilization utilizes a complementary biotinylated LNA capture probe. Location of donor (Cy3 – green) and acceptor (Cy5 – red) fluorophores are indicated. Representative donor-acceptor (top) and smFRET (bottom) traces for PEC-104 in the absence (b) and presence (c) of 0.5 mM Mn²⁺. Probability FRET histogram for each trace is indicated on the right. smFRET histograms of PEC-104 (d), LNA-104 (e) in the presence of only Mg²⁺ (middle panel) and in the presence of both Mg²⁺ and Mn²⁺ (right panel). Cartoon of each construct is indicated on the left. The SD of each FRET value is reported. The total number of traces (N) included in each histogram is indicated on top right corner. Source data are provided with this paper at https://doi.org/10.7302/22513.

At near-physiological concentration of Mg²⁺ (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^13,15. 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 significantly in the absence of Mn²⁺. Upon addition of 0.5 mM Mn²⁺, the riboswitch adopted a more stably docked conformation (Fig. 3c) with the high-FRET population increasing to ~70%, indicating that binding of Mn²⁺ shifts the equilibrium toward the docked state (Fig. 3d and Supplementary Table 4), as previously observed absent the PEC^13,15.

To assess the influence of RNAP on the global dynamics of (un)docking, we assembled the same FRET-labeled 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 the presence of Mg²⁺ ions only, the docked population without RNAP was significantly less populated, with 42% for LNA-104 compared to 53% under the same conditions for PEC-104 (Fig. 3d, e), suggesting that the proximal RNAP promotes the formation of the docked state. Similarly, addition of 0.5 mM Mn²⁺ led to 65% docked state for LNA-104, but a slightly increased 67% for PEC-104 (Fig. 3d, 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 Mg²⁺ 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 Mn²⁺ the fraction of docked state significantly increased to 67 % (Supplementary Fig. 10 and Supplementary Table 4), indicating that the full-length riboswitch remains highly Mn²⁺ 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 Mn²⁺ stabilizes a static docked (SD) conformation in the riboswitch in the presence of a complete P1.1 stem¹³. Further inspection of the smFRET trajectories in our 104 constructs revealed that HMM modeling was best performed using three FRET states (Undocked = U (~0.2), Semi-Docked = D* (~0.45), Docked = D (~0.75)). In the presence of Mg²⁺ 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, b). 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, f), 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 non-obligatory, intermediate between the undocked and undocked states. These findings show that, in the presence of a near-physiological Mg²⁺ 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.

Representative FRET trajectories for PEC-104 in the presence of only Mg²⁺ (a) and in the presence of both Mg²⁺ and Mn²⁺ (c) along with the corresponding Transition Occupancy Density Plots (TODP) and the transition rate constants calculated in the absence (b) and in the presence (d) of Mn²⁺, See also Supplementary Figs. 5 and 6. e–h Representative FRET trajectories for LNA-104 in the presence of only Mg²⁺ (e) and in the presence of both Mg²⁺ and Mn²⁺ (g) along with the corresponding Transition Occupancy Density Plots (TODP) and the transition rate constants calculated in the absence (f) and in the presence (h) of Mn²⁺, See also Supplementary Figs. 8 and 9. TODPs represent dynamic traces as “off-diagonal” and the static traces as “on-diagonal” contour, where the color scale shows the prevalence of each population. The riboswitch conformational state corresponding to each FRET value is indicated. U = Undocked, D* = Pre-Docked, D = Docked. Source data are provided with this paper at https://doi.org/10.7302/22513.

As expected from the FRET population histograms, addition of Mn²⁺ increased the proportion of the on-diagonal SD population in both PEC-104 (from ~20% to ~35%; Fig. 4c, d) 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 Mn²⁺, while the transition from D to U became slower (2.41 s^−1 and 1.67 s^−1 in the absence and presence of Mn²⁺, respectively). Binding of Mn²⁺ significantly 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 Mn²⁺ (2.12 s^−1 compared to 1.76 s^−1 with only Mg²⁺), whereas the transition from D to D* became slower (1.51 s^−1 compared to 1.96 s^−1 with only Mg²⁺), further supporting that the D* state represents a semi-docked state on route to Mn²⁺-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 Mn²⁺, 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 Mn²⁺, 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 riboswitch as observed in the preQ₁-sensing riboswitch PEC from B. subtilis²¹.

As a further control, we analyzed the FRET transition rate constants in the extended LNA-114 construct in the absence and presence of Mn²⁺ ion (Supplementary Fig. 11). With a now-fully available P1.1 switch helix, the riboswitch behaved similarly to LNA-104, exhibiting a significant proportion of SU (42%) and SD (40%) in both the absence and presence of Mn²⁺, 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 semi-docked conformational state D* may depend on an only partially folded P1.1. Even with three-state HMM fits, 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 P1.1¹³, with k_undock becoming slower upon ligand binding. Similarly, k_{undock,D*→U} in the presence of Mn²⁺ was significantly 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 significantly detected in the earlier study featuring the stabilized P1.1¹³.

To further test whether the observed conformational changes are due to stabilization of switching helix P1.1 in PEC-104, we directly surveyed P1.1 folding using smFRET. To measure dynamic distances along the P1.1 helical axis, we employed stepwise transcription to fluorescently label the 5’end of the transcript with Cy3 and incorporate an azido-modified NTP at position 11 (Supplementary Fig. 1), which subsequently was labeled using click-chemistry with DBCO-Cy5^22,55. 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 Mg²⁺ 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 Mn²⁺, 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 Mn²⁺, 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, e). This finding 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 fluoride-sensing riboswitch from B. cereus⁵⁶, or long-lived steric clashes with the nearby RNAP surface.

Taken together, our single-molecule kinetic analyses reveal a previously unobserved semi-docked, pre-folded 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 ligand-sensing 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^44,57,58. Because folding of P1.1 within the RNAP exit channel could act similarly to a Class I transcriptional pause, we next evaluated RNAP pausing efficiency in the presence of NusA (Fig. 5a). Indeed, in the presence of Mg²⁺ 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 Mn²⁺ 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 findings suggest that the ligand-bound riboswitch with its stabilized P1.1 helix supersedes NusA-mediated enhancement of G104 pausing efficiency, as also seen for the fluoride-sensing riboswitch²³, suggesting redundant action by Mn²⁺ and NusA.

a Representative denaturing gel showing the RNAP pauses during the transcription of the Mn²⁺ riboswitch in the presence of 100 nM NusA transcription factor. 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 (−Mn²⁺) and presence (+Mn²⁺) of 0.5 mM Mn²⁺. The chase lanes (Ch) were taken at the end of the time course after an additional incubation with 500 µM rNTPs for 5 min. Unprocessed images are provided in the Supplementary Information. b Fraction of complexes at the G104 pause in the presence of NusA transcription factor as a function of the transcription time in the absence (blue) and presence (orange) of 0.5 mM Mn²⁺. The reported errors are the SD of the mean from n = 2 independent replicates. c Fold-change enhancement of G104 pause half-life in the WT, MP1.1 and MP1.1cp riboswitch variants obtained in the presence of 100 nM NusA factor relative to the same condition in the absence of the transcription factor. Experiments were performed using 25 µM rNTPs in the absence (blue) and presence (orange) of 0.5 mM Mn²⁺. Error bars are the SD of the mean from n = 2 independent replicates. The statistical significance of differences was determined using the two-tailed Student’s t-test (***p < 0.01, **p < 0.05, *p < 0.1). d Single-molecule setup for studying colocalization of NusA factor labeled with a single Cy5 fluorescent dye (NusA-Cy5) with a PEC. Immobilization of the PEC on the microscope slide occurs through the biotin-streptavidin roadblock. Repeated bindings of NusA-Cy5 are monitored through direct excitation of the Cy5 fluorescent dye. Representative single-molecule trajectories showing the binding of NusA-Cy5 (pink) to PEC-104 recorded in the absence (e) or presence (f) of 0.5 mM Mn²⁺ added co-transcriptionally. Hidden Markov modeling (HMM) is indicated on the top of each trace. A.U. arbitrary unit. Plots displaying the cumulative unbound (g) and bound (h) dwell times of NusA-Cy5 in the absence (blue) and presence (orange) of 0.5 mM Mn²⁺ in the context of PEC-104. The binding (k_on) and dissociation (k_off) rate constants of NusA are indicated. The reported errors are the error of the fit. Total number of molecules analyzed for each condition is as follow: (−) Mn²⁺ = 302; (+) Mn²⁺ = 294. i Overall binding rate constants (k_on) of NusA-Cy5 in the context of PEC-104 WT, MP1.1 and MP1.1cp constructs determined in the absence (blue) and presence (orange) of 0.5 mM Mn²⁺. Error bars are the SD of the mean from n = 2 independent replicates. The statistical significance of differences was determined using the two-tailed Student’s t-test (***p < 0.01, **p < 0.05, *p < 0.1). See also Supplementary Fig. 13. Total number of molecules analyzed for each condition is as follow: PEC-104 WT (−) Mn²⁺ = 302; PEC-104 WT (+) Mn²⁺ = 294; PEC-104 MP1.1 (−) Mn²⁺ = 232; PEC-104 MP1.1 (+) Mn²⁺ = 303; PEC-104 MP1.1cp (−) Mn²⁺ = 276; PEC-104 MP1.1cp (+) Mn²⁺ = 301. Source data are provided with this paper at https://doi.org/10.7302/22513.

Because we identified 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 Mn²⁺ on pausing efficiency (Fig. 1d), the mutant allowed NusA to retain its ability to extend pausing independent of Mn²⁺ (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 Mg²⁺ only was lost upon addition of Mn²⁺ (~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 Mn²⁺ 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 Mn²⁺ needed to modulate transcription termination and readthrough (Supplementary Fig. 14a, 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^17,23,51. In further support, an in vitro termination assay performed in either the absence or presence of NusA revealed a small, but significant 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 Mn²⁺ 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–f)²³. In the presence of Mg²⁺ only, NusA bound with two rate constants of 2.10 ± 0.01 × 10⁷ M^−1 s^−1 (k_NusA-on,fast; fraction 25%) and 3.06 ± 0.01 × 10⁶ 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, h 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)²³, 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 Mn²⁺, the cumulative unbound dwell times could only be fitted with a single-exponential k_NusA-on of 3.35 ± 0.01 × 10⁶ M^−1 s^−1 (Fig. 5g), similar to the slower binding rate constant in Mg²⁺ only, indicating that Mn²⁺ 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 Mn²⁺ 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 Mn²⁺ 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 co-transcriptional folding, we next surveyed NusA binding during real-time transcription of the riboswitch^23,59 (Supplementary Fig. 16). In Mg²⁺ only, and absence of Mn²⁺, 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 Mn²⁺ is added co-transcriptionally, a slight, but significant decrease of the bound time is observed (~1.2 s; Supplementary Fig. 16c), indicating that co-transcriptional 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 Mn²⁺-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, Mn²⁺ 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)⁶⁰ it is expected to be almost continuously bound to an active elongation complex (EC) in vivo⁶¹ 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 Mn²⁺ (Fig. 6a).

a smFRET experimental setup. Paused Elongation Complex (PEC) immobilization utilizes biotinylated E. coli RNAP. Location of donor (Cy3 – green) and acceptor (Cy5 – red) fluorophores are indicated. Representative donor-acceptor (top) and smFRET (bottom) traces for NusA-Bound PEC-104 in the absence (b) and presence (c) of 0.5 mM Mn²⁺. Probability FRET histogram for each trace is indicated on the right. SmFRET histograms of NusA-bound PEC-104 in the presence of only Mg²⁺ (d) along with the corresponding Transition Occupancy Density Plots (TODP) and the determined transition rate constants (e). SmFRET histograms of NusA-bound PEC-104 in the presence of both Mg²⁺ and Mn²⁺ (f) along with the corresponding Transition Occupancy Density Plots (TODP) and the determined transition rate constants (g). The SD of each FRET value is reported. The total number of traces (N) included in each histogram is indicated on top right corner. See also Supplementary Figs. 16 and 17. TODPs represent dynamic traces as “off-diagonal” and the static traces as “on-diagonal” contour, where the color scale shows the prevalence of each population. The riboswitch conformational state corresponding to each FRET value is indicated. U = Undocked, D* = Pre-Docked, D = Docked. Source data are provided with this paper at https://doi.org/10.7302/22513.

Similar to the traces collected for PEC-104 in the absence of Mn²⁺, i.e., in the absence of any divalent ion or with Mg²⁺ 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 Mn²⁺ with NusA bound the cumulative FRET histograms could be best fitted with three independent Gaussian peaks, corresponding to the three conformational states (U, D* and D; Fig. 6d, f). Moreover, upon Mn²⁺ 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 semi-docked state.

Kinetic analysis of the transitions between the different conformational states further confirmed 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 Mg²⁺, 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*} significantly 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 Mn²⁺, 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 Mn²⁺-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 significantly slower in the presence of both Mn²⁺ 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), confirming 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, b).

a Single-molecule setup for studying colocalization of NusA factor labeled with a single Cy5 fluorescent dye (NusA-Cy5) and SiM-KARTS probe labeled with a single Cy3 fluorescent dye with a PEC. Immobilization of the PEC on the microscope slide occurs through the biotin-streptavidin roadblock. Repeated bindings of NusA and SiM-KARTS probe are monitored through direct excitation of the Cy3 and Cy5 fluorescent dyes. b Representative single-molecule trajectories showing the binding of NusA-Cy5 (pink) and SiM-KARTS probe (green) to single PEC-104 (blue). A.U., arbitrary unit. Plots displaying the cumulative colocalization time NusA-Cy5 and the SiM-KARTS probe in the absence (blue) and presence (orange) of 0.5 mM Mn²⁺ in the context of PEC-104_WT (c) and PEC-104_MP1 (d). The colocalization dwell times are indicated. The reported errors are the errors calculated from bootstrapping of all dwell times collected. Total number of colocalization events analyzed for each condition is as follow: PEC-104_WT (−) Mn²⁺ = 232; PEC-104_WT (+) Mn²⁺ = 117; PEC-104_MP1 (−) Mn²⁺ = 124; PEC-104_MP1 (+) Mn²⁺ = 91. Source data are provided with this paper at https://doi.org/10.7302/22513.

In the presence of only Mg²⁺, 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 Mn²⁺, this colocalization dwell time significantly 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 Mn²⁺, NusA and RNAP all act; in return, P1.1 integrates these disparate signals as it modulates gene expression.