An anionic ligand snap-locks a long-range interaction in a magnesium-folded riboswitch

The archetypical transcriptional crcB fluoride riboswitch from Bacillus cereus is an intricately structured non-coding RNA element enhancing gene expression in response to toxic levels of fluoride. Here, we used single molecule FRET to uncover three dynamically interconverting conformations appearing along the transcription process: two distinct undocked states and one pseudoknotted docked state. We find that the fluoride anion specifically snap-locks the magnesium-induced, dynamically docked state. The long-range, nesting, single base pair A40-U48 acts as the main linchpin, rather than the multiple base pairs comprising the pseudoknot. We observe that the proximally paused RNA polymerase further fine-tunes the free energy to promote riboswitch docking. Finally, we show that fluoride binding at short transcript lengths is an early step toward partitioning folding into the docked conformation. These results reveal how the anionic fluoride ion cooperates with the magnesium-associated RNA to govern regulation of downstream genes needed for fluoride detoxification of the cell.


Supplementary Note 1: Experiments with an alternative donor labeling position
To validate our findings, we further performed smFRET experiments at RNA lengths of 48 and 58 with a different donor labeling position ( Supplementary Fig. 9a). In this design, the donor (DY547) is placed at A39, rather than U33 as in the previous construct; the acceptor position remains the same at the 5'-end. When referring to this new design, we place "n" after the number designating a particular RNA species (EC58n and RNA58n, for example).
For the isolated RNA58n, we observed only one FRET state and no changes upon addition of Mg 2+ and F -( Supplementary Fig. 9b,d). The FRET histograms were best fitted with a single Gaussian peak centered around EFRET  0.9 in the absence and presence of Mg 2+ and F -( Supplementary Fig. 9d). In contrast, the elongation complex EC58n exhibits an additional mid-FRET state in the absence of Mg 2+ and F -, which becomes more prominent in the presence of Mg 2+ and F - (Supplementary Fig. 9c,e). Since Ffavors the docked pseudoknot conformation of the riboswitch, the additional peak in EC58n presumably results from pseudoknot formation in the presence of RNAP. To ensure that the mid-FRET state represents the docked conformation in this new design, we performed control experiments in which the pseudoknot was disrupted either by mutation (U45A/C46U) 1 or with the L1 block sequence ( Supplementary Fig. 10). Both of these alterations depress the population of the mid-FRET state and abolish its response to Mg 2+ and F -, confirming that the mid-FRET state represents the docked conformation.
Having noticed that the dynamics are very slow at this transcript length for constructs with the original labeling scheme, we performed smFRET measurements with an exposure time of 400 ms. For EC58n, the FRET histograms were best fitted with two Gaussian peaks centered around EFRET  0.55 and 0.9 and no effect was observed upon addition of Fin the absence of Mg 2+ ( Supplementary Fig. 11c). In the presence of 5 mM Mg 2+ , the population of the mid-FRET state was 20%, which increased to 45% in the presence of 0.5 mM F -. These results are in agreement with those shown in the main text for the corresponding RNA designs with the original labeling scheme (Fig. 4), indicating that Fbinds only in the presence of Mg 2+ and induces folding into the docked conformation.
We observed both dynamic and static traces as shown by diagonal and off-diagonal features in TODPs, respectively ( Supplementary Fig. 11c,d). In the absence of Mg 2+ , a static population is observed in the high FRET sate, which on addition of Fremains almost unchanged. However, in the presence of Mg 2+ Faddition leads to a decrease in the static high-FRET population and a concomitant increase in the dynamic population, as well as a small static population in the mid-FRET state at high Fconcentration ( Supplementary Fig. 11d). We observed very slow dynamics for this RNA design ( Supplementary Fig. 11e), where the rate constants are observed as kdock  0.06 s −1 and kundock  0.062 s −1 in the absence of Mg 2+ and almost no change was observed on addition of F -. In the presence of 5 mM Mg 2+ , the rates are observed as kdock  0.03 s −1 kundock  0.032 s −1 , which on addition of 0.1 mM Fchanges to kdock  0.04 s −1 kundock  0.025 s −1 . All these results support that the docked state is stabilized in the presence of F -.
Because the intrinsic dynamics of the aptamer are fast (discussed in the main text for RNA64), the comparably slow dynamics of EC58n point towards involvement of other factors.
Additionally, the mid-FRET (docked) state was observed in EC58n but not RNA58n ( Supplementary Fig. 9). These additional factors could include interactions of the riboswitch with either the whole transcription machinery or with only the DNA template or RNAP. To differentiate between these possibilities, we performed smFRET experiment on RNA58n in the presence of only the DNA template, where the riboswitch was immobilized through a biotinylated nontemplate DNA (ntDNA) strand. Under these conditions, we observed only the high-FRET state ( Supplementary Fig. 12), similar to what was seen for isolated RNA58n (Supplementary Fig. 9d).
This clearly indicates that the docked state observed in EC58n appears due to the presence of RNAP; potential interactions of the riboswitch with RNAP are discussed in the main text and the following section.
We observed mainly the high-FRET state for isolated RNA48n as well as EC48n and did not observe any detectable change upon addition of Mg 2+ and F - (Supplementary Fig. 13). Based on these results and those discussed in the main text for RNA48 and EC48, we conclude that this state is prevalent before the riboswitch has been fully transcribed, and we term it the precursor state.

Supplementary Note 2: Interaction of the riboswitch with RNAP at transcript length 58
We first tested the hypothesis that the riboswitch interacts with RNAP by performing