IntroductionRNA begins to fold cotranscriptionally as it is synthesized by an RNA polymerase (RNAP)1,2,3. Because base pair formation occurs ~3 orders of magnitude faster than nucleotide addition, nascent RNA structures can begin to fold as RNA emerges from RNAP and, in some cases, within the RNA exit channel of RNAP4,5. Consequently, cotranscriptional RNA folding pathways typically comprise a sequence of folding intermediates, some of which may be transient structures that do not persist within the native structure of the full-length RNA6,7,8,9,10,11.In the past decade, biochemical and biophysical methods for measuring cotranscriptional RNA structure and folding have begun to enable the detection of RNA folding intermediates and the reconstruction of RNA folding pathways from data9,10,11,12,13,14,15,16,17,18,19,20. Among the experimental approaches that have been developed, cotranscriptional RNA structure probing applies high-throughput RNA chemical probing to measure the structure of nascent RNA. Existing cotranscriptional RNA structure probing methods capture RNA folding intermediates by systematically arresting RNAP at each position of a DNA template so that cotranscriptionally folded RNA is displayed from a static transcription elongation complex (TEC) and can be chemically probed9,10,13,20. This strategy can be used to measure how the structure of an RNA molecule changes as it emerges from an RNAP and has been applied to several riboswitches and non-coding RNAs9,10,13,19,20,21,22. While systematic cotranscriptional RNA structure probing experiments can detect structural rearrangements that occur as a nascent transcript grows longer, they do not assess whether one intermediate structure can rearrange into another structure cotranscriptionally because each reactivity profile is an end-point measurement. Consequently, while systematic nascent RNA structure probing experiments measure the structure of cotranscriptionally folded RNA, they do not directly measure cotranscriptional RNA folding. It is therefore possible that equilibration of a nascent transcript within a static TEC prior to chemical probing could cause the formation of a non-native structure that is not a true folding intermediate and which, in some cases, may not be able to rearrange into downstream native structures. This limitation of cotranscriptional RNA structure probing is partially addressed by the use of single-molecule force spectroscopy12,18 and single-molecule FRET14,15,16,17 to measure cotranscriptional RNA folding continuously with high temporal resolution. However, the ability to measure the cotranscriptional rearrangement of RNA structures by chemical probing would provide a means for assessing the validity of cotranscriptional RNA folding events that captures structural information for each nucleotide of the target RNA.To facilitate the detection of cotranscriptional RNA folding events by chemical probing, we have developed linked-multipoint Transcription Elongation Complex RNA structure probing (TECprobe-LM). TECprobe-LM uses the SHAPE-MaP-based23 TECprobe platform20 that we developed previously to directly assess whether the reactivity profiles of two RNA folding intermediate populations are linked by a cotranscriptional folding event. In a TECprobe-LM experiment, aliquots of an E. coli RNAP in vitro transcription reaction are removed for chemical probing after RNAP has arrested at a photolabile NPOM-caged-dT24 stall site25 and after RNAP has arrested at a downstream biotin-streptavidin roadblock following removal of the NPOM-cage by irradiation with 365 nm UV light (Fig. 1). In this way, the cotranscriptional conversion of one population of RNA structures into a second population of structures is directly measured by chemical probing. To validate this approach, we used TECprobe-LM to visualize the rearrangement of a non-native intermediate hairpin into the native E. coli signal recognition particle RNA, folding of the C. beijerinckii pfl ZTP riboswitch aptamer and expression platform, and folding of the B. cereus crcB fluoride riboswitch aptamer and expression platform. All cotranscriptional folding transitions were detected by TECprobe-LM, and the observed folding intermediates agree with measurements made by systematic cotranscriptional RNA structure probing experiments in all but one case. TECprobe-LM provided a higher resolution view of pfl ZTP aptamer pseudoknot folding than previous methods, which revealed 1) that pseudoknot folding becomes possible over a single nucleotide addition cycle from +102 to +103, and 2) that, given the ~14 nt footprint of E.coli RNAP26, the RNA exit channel can accommodate at least 3 pseudoknot base pairs. The primary limitation of TECprobe-LM is that, like other cotranscriptional RNA structure probing methods, the representation of transcripts in a sequencing library can depend on the efficiency of an ssRNA ligation that is prone to sequence and structure biases. Nonetheless, it was possible to collect high-quality data for all samples described in this work, including one sample for which this ligation was particularly inefficient. Together, our findings establish TECprobe-LM as a strategy for measuring the cotranscriptional rearrangement of RNA structures using high-throughput RNA chemical probing.Fig. 1: Overview of TECprobe-LM.Single-round transcription is initiated on template DNA that contains NPOM-caged-dT and terminal biotin-streptavidin roadblocks. In the initial phase of transcription, RNAP arrests at the NPOM-caged-dT roadblock. The pre-wash sample is removed from the reaction and chemically probed, and the remaining transcription reaction is washed to remove NTPs. The NPOM cage is removed by irradiation with 365 nm UV light (10 mW/cm2 for 3 min) and the pre-chase sample is removed from the reaction and chemically probed. Upon addition of NTPs, RNAP transcribes to the terminal biotin-streptavidin roadblock and the post-chase sample is removed and chemically probed. Figure elements depicting RNA polymerase and nascent RNA were adapted from Szyjka and Strobel, Observation of coordinated RNA folding events by systematic cotranscriptional RNA structure probing20. RNAP, RNA polymerase; SRP, signal recognition particle; SAv, streptavidin; BzCN, benzoyl cyanide.Full size imageResultsOverview of the linked multi-point cotranscriptional RNA structure probing strategyLinked multi-point cotranscriptional RNA structure probing directly assesses whether one population of nascent RNA structures can cotranscriptionally rearrange into a second population of structures. This is accomplished by removing aliquots of an E. coli RNAP in vitro transcription reaction for chemical probing as RNAP is moved to specific locations within the DNA template (Fig. 1). In the simplest implementation of TECprobe-LM described in this work, RNAP first transcribes to one nucleotide upstream of an NPOM-caged-dT modification in the template DNA strand, which was shown by Nadon et al. to function as a photoreversible transcription roadblock25. A ‘pre-wash’ aliquot of the transcription reaction is then removed for chemical probing and the arrested TECs are washed extensively to remove excess NTPs. The NPOM-photocage is then removed by exposure to 365 nm UV light (10 mW/cm2 for 3 min) and a ‘pre-chase’ aliquot of the transcription reaction is removed for chemical probing. NTPs are then added to the reaction so that RNAP transcribes to a terminal biotin-streptavidin roadblock and a ‘post-chase’ aliquot of the transcription reaction is removed for chemical probing. The pre-wash and pre-chase samples can be compared to confirm that the population of RNA structures that existed after RNAP first stalled at the NPOM-caged-dT site did not change when the arrested TECs were washed. The pre-chase and post-chase samples can be compared to assess how the population of RNA structures changed when RNAP transcribed downstream. In this way, the linked multi-point strategy for cotranscriptional RNA structure probing directly assesses whether a population of nascent RNA structures at an earlier point within a cotranscriptional folding pathway can rearrange into a second population of structures at a later point in the folding pathway. This measurement is distinct from existing cotranscriptional RNA chemical probing methods that measure nascent RNA structure in the context of static TECs, which do not directly assess whether structures that are observed early in a folding pathway can rearrange into structures that are observed later in the folding pathway.Rearrangement of the 4.5S SRP RNA intermediate hairpinWong et al. previously showed that the E. coli signal recognition particle (SRP) RNA can form a non-native structure prior to folding of the native SRP RNA structure8. This model was later refined by Watters, Strobel et al. using Cotranscriptional SHAPE-Seq9. While these prior studies detected a non-native intermediate hairpin, they did not directly assess whether the intermediate hairpin could refold into the mature SRP RNA structure. Fukuda et al. later detected cotranscriptional refolding of the intermediate hairpin using single-molecule force spectroscopy18 and Yu et al. identified efficient mechanisms by which this structural transition could occur cotranscriptionally19. To visualize rearrangement of the SRP RNA intermediate hairpin by high-throughput RNA chemical probing, we performed a TECprobe-LM experiment in which RNAP was first positioned at +127, which precedes intermediate hairpin rearrangement, and then chased to +161 at which point the native SRP RNA structure is expected to have folded (Fig. 2a).Fig. 2: Cotranscriptional folding of the E. coli SRP RNA.a Secondary structures of the E. coli SRP RNA folding intermediates that were assessed by TECprobe-LM colored by reactivity. Sequence within the RNAP footprint is not shown. b Transcript length distribution for the pre-wash, pre-chase, and post-chase samples. Traces are the average of n = 2 replicates. c Comparison of reactivity profiles of pre-wash and pre-chase samples (upper plot) and of pre-chase and post-chase samples (lower plot). Solid lines are the average of n = 2 replicates and reactivity values for individual replicates are shown as points. Source data are provided as a Source Data file. RNAP, RNA polymerase; PK, pseudoknot; SAv, streptavidin; BzCN, benzoyl cyanide.Full size imageIn the pre-wash and pre-chase samples, ~98% of aligned reads mapped to transcripts upstream of the NPOM-caged-dT modification at +128, ~75% mapped to +127, and ~5% mapped to +126 (Fig. 2b, Supplementary Fig. 1a). In the post-chase sample, the fraction of aligned reads that mapped to transcripts beyond +127 increased from ~2% to ~57%, and ~40% mapped to the biotin-streptavidin roadblock enrichment sites from +158 to +162 (Fig. 2b, Supplementary Fig. 1a). The presence of a roadblock-independent enrichment site at +142 suggests that RNAP is prone to arresting at this position (Fig. 2b, Supplementary Fig. 1a). Both the 142 nt transcript and biotin-streptavidin-enriched transcripts were observed by denaturing PAGE (Supplementary Fig. 2a). However, in contrast to the read distribution observed by TECprobe-LM, the 142 nt transcript was more abundant than transcripts that were enriched by the biotin-streptavidin roadblock when assessed by gel electrophoresis. This difference is most likely caused by transcript-specific variation in the efficiency of 3’ adapter ligation skewing the representation of nascent transcripts in the sequencing library13.Washing the roadblocked TECs to remove NTPs did not perturb RNA structure (Fig. 2c, upper plot). Several known elements of the SRP RNA structure were observed when RNAP was positioned at +127: i) the 5’ leader hairpin and linker were detected as elevated reactivity at G9, U10, and C12 and at nts 24–26, respectively, ii) the intermediate hairpin loop was detected as elevated reactivity at nucleotides 33–41, and iii) the apical loop of the native SRP RNA structure, upstream segment of bulge D, and downstream segment of bulge B were detected as elevated reactivity at nucleotides within each of these regions of the SRP RNA hairpin (Fig. 2a, c, lower plot). Upon transcription from +127 to +161, nucleotides within the intermediate hairpin loop became non-reactive while the reactivity of flexible nucleotides in the native SRP RNA structure and the 5’ leader persisted, indicating that the non-native intermediate hairpin had refolded (Fig. 2a, c, lower plot). The reactivity profiles obtained using TECprobe-LM agreed with end-point profiles collected using variable length TECprobe (TECprobe-VL) except that the leader hairpin and intermediate hairpin loops were more reactive in TECprobe-LM experiments (Supplementary Fig. 3).C. beijerinckii pfl ZTP riboswitch aptamer foldingThe ZTP aptamer comprises two sub-domains that form a long-range pseudoknot27 (Fig. 3a, 120 nt transcript). ZMP binding establishes a contiguous helical stack between P3 and the pseudoknot (PK) which, in the C. beijerinckii pfl ZTP aptamer, blocks nucleation of the terminator hairpin11,28,29,30,31. To visualize pfl aptamer folding, we initially performed a TECprobe-LM experiment in which RNAP was first positioned at +102, which precedes pseudoknot and P3 subdomain folding, and then chased to +120 at which point the pseudoknot and P3 are folded and ZMP can bind.Fig. 3: Cotranscriptional folding of the C. beijerinckii pfl ZTP riboswitch aptamer.a Secondary structures of the C. beijerinckii pfl ZTP riboswitch aptamer folding intermediates that were assessed by TECprobe-LM colored by reactivity. b Transcript length distribution for the pre-UV, pre-chase, and post-chase samples. Traces are the average of n = 2 replicates. c Difference in reactivity (1 mM ZMP − 0 mM ZMP) observed for the pre-UV, pre-chase, and post-chase samples. Differences were calculated from the average reactivity values shown in (d) and (e). d, e Comparison of reactivity profiles of pre-UV and pre-chase samples (upper plot) and of pre-chase and post-chase samples (lower plot) in the absence (d) and presence (e) of 1 mM ZMP. Solid lines are the average of n = 2 replicates and reactivity values for individual replicates are shown as points. Source data are provided as a Source Data file. RNAP, RNA polymerase; PK, pseudoknot; SAv, streptavidin; BzCN, benzoyl cyanide.Full size imageIn the pre-wash sample, ~91% of aligned reads mapped to transcripts upstream of the NPOM-caged-dT modification at +103, ~60% mapped to +102, and ~4% mapped to +101 (Supplementary Fig. 4a). ~6% of aligned reads mapped to +103, which indicates that RNAP can insert a nucleotide opposite of the NPOM-caged-dT modification in some sequence contexts, including the poly-dT tract of the current construct. In the pre-chase sample, RNAP incorporated 1-2 nucleotides following UV irradiation despite extensive washing to remove excess NTPs (Supplementary Fig. 4a). Comparison of reactivity profiles of these transcripts revealed that translocation of RNAP from +102 to +103 permits the ZTP aptamer pseudoknot to fold (Supplementary Fig. 4b). Consequently, TECs in the pre-chase sample comprised a mixed population of ≥103 nt RNAs, which have formed the pseudoknot, and 102 nt RNAs which have not. In the post-chase sample, the fraction of aligned reads that mapped to transcripts beyond +103 increased from ~2% to >80–90%, and ~80–84% mapped to the biotin-streptavidin roadblock enrichment sites from +116 to +122 (Supplementary Fig. 4a). The transcript distribution observed by sequencing agreed with the distribution observed by denaturing PAGE (Supplementary Fig. 2b).We reasoned that one way to resolve the mixed population that was observed in the pre-chase sample is to favor translocation to ≥ + 103 by increasing the concentration of NTPs used during the initial phase of transcription and washing the roadblocked TECs less extensively. This would enable a three-point TECprobe-LM experiment in which two sequential cotranscriptional RNA folding transitions are observed (Fig. 3a). In this format, the first sample that is collected is referred to as the ‘pre-UV’ sample, and the pre-wash control is omitted for simplicity because washing TECs to remove NTPs did not perturb RNA structure when the experiment was performed using the standard TECprobe-LM format (Supplementary Fig. 4c). Increasing NTP concentration and reducing the wash volume caused ~86% of TECs that were arrested at the NPOM-caged-dT stall site to transcribe to positions +103 to +106 upon release of the NPOM cage (Fig. 3b, Supplementary Fig. 1b). The reactivity profiles of the resulting 103 to 106 nt transcripts were virtually identical (Supplementary Fig. 4d). As observed in the initial experiment described above, ~80% of aligned reads mapped to the biotin-streptavidin enrichment sites from +116 to +122 in the post-chase sample (Fig. 3b, Supplementary Fig. 1b).As expected, ZMP binding was not detected when RNAP was positioned at +102 because the aptamer had not yet folded (Fig. 3c). In both the absence and presence of ZMP, transcription to +104 caused the reactivity of nucleotides 24–26, A88, C89, and G91 to decrease upon pseudoknot folding (Fig. 3a, d, e, upper plots). ZMP binding was still not detected because the P3 hairpin had not yet folded (Fig. 3c). As described above, the signatures of pseudoknot folding that were observed at +104 were also observed at +103 (Supplementary Fig. 4b, d). This shows that transcription from +102 to +103 exposes enough of the nascent transcript to permit pseudoknot folding and that, given the ~14 nt footprint of RNAP on RNA26, the pseudoknot can stably fold when four of the five L3 nucleotides that form pseudoknot base pairs are present in the RNA exit channel. The observation of pseudoknot folding as reduced reactivity at A24, C25, G26, C89, and G91 upon formation of the A24:U92, C25:G91, G26:C90, and G27:C89 base pairs suggests that at least three pseudoknot base pairs are accommodated in the RNA exit channel of RNAP (Fig. 3a and Supplementary Fig. 5a). The G27:C89 base pair forms outside of the RNA exit channel and it is not clear whether the G23:C93 base pair can form at +103 because both G23 and C93 exhibited low reactivity at +102, before pseudoknot folding can occur. A previous analysis of refolded intermediate transcripts using SHAPE-Seq showed that, in the absence of RNAP, the pfl ZTP aptamer pseudoknot becomes stable once all five base pairs can form11 (Supplementary Fig. 5b). This suggests that pseudoknot folding at +103 is facilitated either by interactions between nascent RNA and the RNA exit channel that stabilize a four base pair pseudoknot32,33 or by formation of the G23:C93 base pair. We cannot exclude the possibility that pseudoknot folding at +103 could be facilitated by nascent RNA structure driving RNAP from a post-translocated state into a hyper-translocated state, which would expose additional RNA sequence. However, the observation that ~95% of TECs at +103 in the pre-chase samples transcribed downstream upon addition of NTPs indicates that if RNA structure-dependent hyper-translocation does occur it must not be persistent (Fig. 3b).Chasing RNAP to +120 in the absence of ZMP caused the reactivity of nucleotides 84-86 to decrease as P3 folded (Fig. 3a, d, lower plot). When RNAP was chased to +120 in the presence of ZMP, P3 folding was observed in coordination with expected ZMP-dependent reactivity changes including: i) decreased reactivity in P1 due to stabilization of non-canonical base pairs, ii) decreased reactivity of A34 and A38 in L2 which corresponds to formation of a conserved A-minor motif, and iii) decreased reactivity at U94 and G95 in L3 which hydrogen bond and stack with the Z nucleobase, respectively11,20 (Fig. 3a, c, e, lower plot). The reactivity profiles obtained using TECprobe-LM agreed with profiles collected using TECprobe-VL with two exceptions: First, the reactivity of nucleotides immediately upstream of the RNAP footprint in the 102 nt transcript was higher in the TECprobe-LM profile (Supplementary Fig. 6a, b). Second, the TECprobe-LM profiles of the 104 nt transcript more closely matched the TECprobe-VL profiles of the 110 nt transcript, in which the pseudoknot has fully folded, than the TECprobe-VL profiles of the 104 nt transcript (Supplementary Fig. 6). Both exceptions are most likely caused by backtracking that can occur when RNAP collides with a biotin-streptavidin roadblock during the TECprobe-VL procedure13, which shifts the RNAP footprint upstream and prevents the formation of RNA structures that would otherwise fold if the nascent transcript 3’ end were positioned at the RNAP active center.C. beijerinckii pfl ZTP riboswitch terminator foldingTo visualize pfl ZTP riboswitch terminator hairpin folding, we performed a TECprobe-LM experiment in which RNAP was first positioned at +111, at which point P3 can fold while partially in the RNA exit channel of RNAP, and then chased downstream of the termination site to +143 (Fig. 4a, b). In this experimental configuration, ZMP binding occurs cotranscriptionally as the ZTP aptamer emerges from RNAP until ~+123 when the terminator hairpin can nucleate.Fig. 4: Cotranscriptional folding of the C. beijerinckii pfl ZTP riboswitch expression platform.a, b Secondary structures of the C. beijerinckii pfl riboswitch expression platform folding intermediates that were assessed by TECprobe-LM in the absence (a) and presence (b) of 1 mM ZMP colored by reactivity. c Transcript length distribution for the pre-wash, pre-chase, and post-chase samples. Traces are the average of n = 2 replicates. d Difference in reactivity (1 mM ZMP − 0 mM ZMP) observed for the pre-wash, pre-chase, and post-chase samples. Differences were calculated from the average reactivity values shown in (e) and (f). e, f Comparison of reactivity profiles of pre-wash and pre-chase samples (upper plot) and of pre-chase and post-chase samples (lower plot) in the absence (e) and presence (f) of 1 mM ZMP. Solid lines are the average of n = 2 replicates and reactivity values for individual replicates are shown as points. Source data are provided as a Source Data file. RNAP, RNA polymerase; PK, pseudoknot; SAv, streptavidin; BzCN, benzoyl cyanide.Full size imageIn the pre-wash sample, ~97% of aligned reads mapped to transcripts upstream of the NPOM-caged-dT modification at +112, ~62% mapped to +111, and ~4% mapped to +110 (Fig. 4c, Supplementary Fig. 1c). In the pre-chase sample, NPOM-caged-dT-enriched transcripts were observed at +111 and +112, which indicates that RNAP was able to incorporate one additional nucleotide upon release of the NPOM cage in some cases despite extensive washing (Fig. 4c, Supplementary Fig. 1c). The reactivity profiles of the 111 nt and 112 nt transcripts were indistinguishable in the absence of ZMP (Supplementary Fig. 7a, upper plot). In the presence of ZMP, the reactivity of ZMP-responsive nucleotides in the 112 nt transcript decreased to a value between that of the apo aptamer and the ZMP-bound aptamer (Supplementary Fig. 7a, lower plot, b). This indicates that 112 nt transcripts can bind ZMP to some extent, although it is not clear whether the intermediate reactivity values are caused by binding of ZMP to a subset of aptamers or by interconversion between apo and ZMP-bound states. As shown by denaturing PAGE, most TECs that were positioned at +111/ + 112 resumed transcription and yielded terminated and full-length RNA products (Supplementary Fig. 2c). In contrast with this observation, 80% of aligned reads mapped to the biotin-streptavidin roadblock enrichment sites from +67 to +73 (Fig. 5c, Supplementary Fig. 1d). The transcript distribution observed by sequencing agreed with the distribution observed by denaturing PAGE (Supplementary Fig. 2d).Washing the roadblocked TECs to remove NTPs did not perturb RNA structure (Fig. 5e, f, upper plots). As expected, fluoride had no effect on RNA structure when RNAP was positioned at +54 because the aptamer had not yet folded (Fig. 5d). Chasing RNAP from +54 to +71 in the absence of fluoride caused an increase in the reactivity of U11, which caps an extended P3 stack that is stabilized by fluoride binding35 (Fig. 5a, e, lower plot). In previous TECprobe-VL experiments that were performed without fluoride, increased U11 reactivity was observed in coordination with decreased reactivity at A14 and G15 due to pseudoknot folding20 (Supplementary Fig. 9a, b, upper plots). This implies that U11 becomes reactive as other nucleotides within L1 form pseudoknot base pairs in the absence of fluoride. However, in the current TECprobe-LM data, the decrease in reactivity at A14 and G15 that was observed when RNAP transcribes from +54 to +71 was negligible (Fig. 5e, lower plot). It is unlikely that the pseudoknot can fold when RNAP is positioned at +54 because two of six downstream pseudoknot nucleotides are expected to be paired within the RNA-DNA hybrid. Furthermore, the reactivity profiles of the 53 and 52 nt transcripts, in which additional downstream pseudoknot nucleotides are paired within the RNA-DNA hybrid, were identical to that of the 54 nt transcript except that A40 was protected from benzoyl cyanide modification when RNAP was positioned at +52 (Supplementary Fig. 10). This indicates that the approximately constant reactivity observed at A14 and G15 is not due to the pseudoknot folding at +54 and that fluoride-independent pseudoknot folding is not detected in the TECprobe-LM experiment. Comparison of TECprobe-LM and TECprobe-VL reactivity profiles of the 54 nt transcript indicates that the nascent RNA adopts a distinct conformation in each experiment (Supplementary Fig. 9a, c, upper plots). Most notably, nucleotides A14 and G15 within L1 and G28, U30 and A35 in P3 were more reactive in the TECprobe-VL profile than in the TECprobe-LM profile. The latter observation suggests that P3 is stably folded when RNAP is positioned at +54 in the TECprobe-LM experiment but not in the TECprobe-VL experiment. In further support of this interpretation, the TECprobe-LM profile of the 54 nt transcript was similar to the TECprobe-VL profile of the 68 nt transcript, except that U11 was reactive at +68 but not at +54 (Supplementary Fig. 9b, upper plot). This indicates that during the TECprobe-LM experiment, P3 folds earlier than was observed by TECprobe-VL. While the cause of this difference is unclear, the agreement of the TECprobe-LM and TECprobe-VL reactivity profiles of the 71 nt transcript in both ligand conditions indicates that the reactivity differences observed at +54 do not interfere with aptamer folding or fluoride binding (Supplementary Fig. 9a, c, lower plots). The observation that transcription from +54 to +71 causes U11 to become reactive in the absence of fluoride and causes fluoride-dependent stabilization of the pseudoknot in the presence of fluoride indicates that the pseudoknot likely folds transiently until fluoride binds. This is consistent with observations made by single-molecule FRET, in which the crcB fluoride aptamer exists in a dynamic docked state until fluoride binding stabilizes the pseudoknot37.Established fluoride-dependent reactivity changes were observed when RNAP was chased from +54 to +71 in the presence of fluoride, including: i) decreased reactivity at A14, G15, and U45 due to pseudoknot stabilization, ii) decreased reactivity at U38, which forms a reversed Watson-Crick pair with A10 that extends the P3 stack, and at U11, which caps the extended P3 stack, iii) decreased reactivity at A40, which forms the linchpin base pair with U48, iv) decreased reactivity at A49, which stacks with the linchpin base pair, v) increased reactivity at U23 and decreased reactivity at A24 and A25 within J1/3, and vi) increased reactivity at U30, C32, and U33 within the P3 stem and loop20 (Fig. 5b, d, f, lower plot). The reactivity profiles of the 71 nt transcript obtained using TECprobe-LM agreed with the TECprobe-VL profiles except that the reactivity of nucleotides immediately upstream of the RNAP footprint was lower in the TECprobe-LM data (Supplementary Fig. 9a, c, lower plots). This difference is most likely caused by the terminal biotin-streptavidin roadblock inducing more extensive backtracking in the TECprobe-LM experiment than the internal biotin-streptavidin roadblocks in the TECprobe-VL experiment.B. cereus crcB fluoride riboswitch terminator foldingTo visualize crcB fluoride riboswitch terminator hairpin folding, we performed a TECprobe-LM experiment in which RNAP was first positioned at +74, after the fluoride aptamer can fold and fluoride can bind, and then chased downstream of the termination site to +95 (Fig. 6a, b). In the pre-wash sample, 99% of aligned reads mapped to transcripts upstream of the NPOM-caged-dT modification at +75, ~88% mapped to +74, and ~5% mapped to +73 (Fig. 6c, Supplementary Fig. 1e). In the pre-chase sample, ~70% of aligned reads mapped to +74 and ~16% mapped to +73 (Fig. 6c, Supplementary Fig. 1e). In the post-chase sample, ~71–76% of aligned reads mapped to the biotin-streptavidin roadblock enrichment sites from +93 to +96 (Fig. 6c, Supplementary Fig. 1e). Although both terminated and full-length transcripts were detected by denaturing PAGE, terminated transcripts were depleted in the TECprobe-LM sequencing library (Fig. 6c, Supplementary Figs. 1e and 2e). This is most likely caused by inefficient ligation of the 3’ adapter since full-length transcripts face the same reverse transcription barriers as terminated transcripts. Nonetheless, it was possible to assess expression platform folding by comparing the reactivity profiles of transcripts within TECs that were arrested at +74 to transcripts within TECs that were arrested at +95.Fig. 6: Cotranscriptional folding of the B. cereus crcB fluoride riboswitch expression platform.a, b Secondary structures of the B. cereus crcB fluoride riboswitch expression platform folding intermediates that were assessed by TECprobe-LM in the absence (a) and presence (b) of 10 mM NaF colored by reactivity. Sequence within the RNAP footprint is not shown. c Transcript length distribution for the pre-wash, pre-chase, and post-chase samples. Traces are the average of n = 2 replicates. d Difference in reactivity (10 mM NaF − 0 mM NaF) observed for the pre-wash, pre-chase, and post-chase samples. Differences were calculated from the average reactivity values shown in (e) and (f). e, f Comparison of reactivity profiles of pre-wash and pre-chase samples (upper plot) and of pre-chase and post-chase samples (lower plot) in the absence (e) and presence (f) of 10 mM NaF. Solid lines are the average of n = 2 replicates and reactivity values for individual replicates are shown as points. Source data are provided as a Source Data file. RNAP, RNA polymerase; PK, pseudoknot; SAv, streptavidin; BzCN, benzoyl cyanide.Full size imageWashing the roadblocked TECs to remove NTPs did not perturb the structure of the fluoride aptamer but caused a small decrease in the reactivity of nucleotides A58, U60, and A61 that was larger in the presence of fluoride (Fig. 6e, f, upper plots). As expected, fluoride binding was detected when RNAP was positioned at +74 because the fluoride aptamer had folded (Fig. 6d). In both the absence and presence of 10 mM fluoride, chasing RNAP to +95 caused known signatures of terminator hairpin folding20 including: i) increased reactivity at A14 and G15 due to pseudoknot disruption, ii) increased reactivity at nucleotides 38–40 due to disruption of the A10:U38 and A40:U48 pairs, and iii) decreased reactivity in terminator stem nucleotides (Fig. 6a, b, e, f, lower plots). The partial persistence of fluoride-dependent differences between the 0 mM and 10 mM NaF post-chase samples indicates that some fluoride aptamers did remain intact in the presence of the terminator hairpin (Fig. 6d). In contrast, the fluoride-bound aptamer remained mostly intact when TECs were arrested downstream of the terminator from +113 to +116 in previous TECprobe-VL experiments20. As described below, the TECprobe-LM and TECprobe-VL profiles of the 95 nt transcript agreed except that the TECprobe-VL data was noisier due to low sequencing depth because transcripts from +84 to +112 were not enriched. This indicates that the transcription terminator hairpin can fold when RNAP is halted at +95, but that uninterrupted transcription further downstream disfavors disruption of the fluoride-bound aptamer by the terminator hairpin. This implies that the synthesis of downstream RNA or increasing the distance between the riboswitch and RNAP disfavors terminator hairpin folding. Notably, the ability of the fluoride aptamer to block terminator base pair propagation appears to play a minor role in transcription antitermination because the fluoride-bound aptamer delays terminator hairpin nucleation until after RNAP has bypassed the termination site9,13,20.The TECprobe-LM reactivity profiles of the 74 nt transcript agreed with the TECprobe-VL profiles except that the reactivity of nucleotides immediately upstream of the RNAP footprint was higher in the TECprobe-LM data (Supplementary Fig. 11a, c, upper plots). As described above, this is most likely caused by biotin-streptavidin roadblock-induced backtracking shifting the RNAP footprint upstream in the TECprobe-VL experiment13. The TECprobe-LM and TECprobe-VL profiles of the 95 nt transcript agreed except that terminator hairpin stem reactivity was lower in the 0 mM NaF TECprobe-LM data, which may be due to ~82-fold higher sequencing depth yielding a higher quality reactivity profile (Supplementary Fig. 11a, c, lower plots). In support of this interpretation, terminator stem nucleotides were weakly reactive in the TECprobe-VL profile of the 80 nt terminated transcript, which was sequenced ~42.5-times deeper than the 95 nt transcript (Supplementary Fig. 11b).DiscussionTECprobe-LM uses RNA chemical probing to directly assess whether one RNA folding intermediate can cotranscriptionally rearrange into another folding intermediate. The primary advance of the linked-multipoint cotranscriptional RNA structure probing strategy relative to previous approaches is that it assesses whether two RNA folding intermediates are linked by a cotranscriptional folding event, which was not previously possible to measure directly by chemical probing. However, like other cotranscriptional RNA structure probing methods, there is a possibility that nascent RNA structures will equilibrate after transcription has been arrested. Because of this, the linked-multipoint strategy for cotranscriptional RNA chemical probing does not measure true cotranscriptional RNA folding like single-molecule force spectroscopy12,18 and single-molecule FRET experiments14,15,16,17. Nonetheless, the ability to assess whether one population of nascent RNA structures can rearrange into another population of structures using high-throughput RNA structure probing is complementary to these biophysical approaches because it yields structural information for each nucleotide of the transcript.There are several technical limitations that must be considered when designing linked-multipoint cotranscriptional RNA structure probing experiments. First, although many photocaged nucleotides have been incorporated into DNA38,39,40, only NPOM-caged-dT is currently commercially available. For most users, this limits the potential sites that can be used to reversibly halt RNAP. Second, like other cotranscriptional RNA structure probing methods, TECprobe-LM requires the ligation of a sequencing adapter to the nascent RNA 3’ end. The efficiency of this ligation can vary depending on RNA sequence and structure13,41,42. Consequently, the distribution of transcript lengths that is generated by the transcription reaction may not be accurately captured within a sequencing library. The effects of these biases were observed in several ways in the systems assessed here. For example, terminated fluoride riboswitch transcripts were nearly undetectable in TECprobe-LM data but were clearly observed by gel electrophoresis. Similarly, terminated and full-length ZTP riboswitch transcripts were severely underrepresented in TECprobe-LM data but were clearly observed by gel electrophoresis. This limitation cannot currently be circumvented but also did not prevent high-quality chemical probing data from being obtained in the systems that were assessed in this work. Third, the sequence context of the NPOM-caged-dT modification can affect experiment outcomes. For example, TECs that were arrested at the NPOM-caged-dT modification in the SRP RNA DNA template resumed transcription less efficiently than the TECs that were arrested at other NPOM-caged-dT sites. For these reasons, the success of a TECprobe-LM experiment is dependent on the target RNA sequence.In general, the reactivity profiles obtained using TECprobe-LM agreed with end-point cotranscriptional RNA structure probing measurements made by TECprobe-VL. However, in several samples the reactivity of nucleotides that would be immediately outside of the RNA exit channel if the RNA 3’ end were positioned at the RNAP active center was higher when RNAP was arrested at an NPOM-caged-dT roadblock. This observation is consistent with the prior observation that collision of an E. coli TEC with a biotin-streptavidin complex can cause RNAP to backtrack, which would shift the RNAP footprint upstream13. For this reason, it is critical to consider how backtracking can cause uncertainty in the exact transcript length at which a folding transition occurs when using TECprobe-VL data to select sites at which RNAP will be arrested by an NPOM-caged-dT modification. When necessary, candidate roadblocking sites could be assessed quickly and relatively inexpensively by preparing DNA templates that contain an internal amino linker43 at potential roadblocking sites. These templates could then be pooled and used in a small-scale TECprobe-VL experiment that circumvents biotin-streptavidin-induced backtracking. More notably, in the TECprobe-LM reactivity profile of the 54 nt fluoride riboswitch transcript, P3 folded earlier than in the TECprobe-VL experiment and the reactivity of L1 did not change upon pseudoknot folding. While the cause of this difference is not clear, notable differences in the procedures include the use of solution (VL) vs solid-phase (LM) transcription, and the inclusion of free streptavidin in the transcription reaction (VL). Nonetheless, the TECprobe-LM and TECprobe-VL datasets converged upon the same structure once RNAP transcribed downstream to +71 despite having different starting structures. We anticipate that TECprobe-VL data will typically be used to inform the design of TECprobe-LM experiments, which will facilitate comparisons between measurements made by the two methods as we have presented here.The design of every TECprobe-LM experiment described above was informed by TECprobe-VL experiments that had mapped an end-point structure for every intermediate transcript of the target RNA. When applied in this way, TECprobe-LM assesses whether structural transitions that are inferred from end-point structures can occur cotranscriptionally. While the use of TECprobe-VL data to inform the design of TECprobe-LM experiments is advantageous, it is not essential. For example, computationally predicted cotranscriptional RNA folding transitions44,45,46,47 could be used to inform the selection of transcription roadblocking sites in a TECprobe-LM experiment. It would likely remain advantageous to measure end-point structures of the targeted folding intermediates using TECprobe-VL to ensure that the TECprobe-LM experiment will assess the desired cotranscriptional folding event. However, this could be assessed in a small-scale cotranscriptional RNA structure probing experiment that only targets the transcripts of interest.The precision with which NPOM-caged-dT halts E. coli RNAP enabled the observation of pfl ZTP riboswitch pseudoknot folding as RNAP transcribes from +102 to +103. This abrupt folding transition is particularly notable because four of five L3 nucleotides that participate in pseudoknot base pairs are located in the RNA exit channel of RNAP when pseudoknot folding is observed. Based on changes in reactivity that occur upon pseudoknot formation, the RNA exit channel of RNAP can accommodate at least three pseudoknot base pairs in this sequence context. While previous studies have measured the formation of dsRNA duplexes in the RNA exit channel in E. coli hairpin-stabilized pause TECs48,49 and pre-termination complexes50, the extent to which pseudoknotted dsRNA base pairs can fold within the RNA exit channel is not well-defined. Widom et al. previously used molecular dynamics simulations and cross-linking experiments to determine that 1-2 base pairs of the B. subtilis preQ1 riboswitch can fold within the RNA exit channel of E. coli RNAP, which stabilizes a transcription pause51, and Chauvier and Porta et al. confirmed this prediction using cryo-EM52. While it is not clear whether folding of the pfl ZTP riboswitch pseudoknot within the RNA exit channel occurs during continuous transcription, our findings indicate 1) that the RNA exit channel of E. coli RNAP can accommodate at least three pseudoknotted dsRNA base pairs, and 2) that, in the context of RNAP, stable folding of the pfl ZTP riboswitch pseudoknot requires the formation of at least four of five pseudoknot base pairs, which can occur when RNAP is positioned at +103. These observations could not have been made using end-point cotranscriptional RNA structure probing measurements because it was necessary to observe that +103 TECs remain active, which indicates that the formation of pseudoknot base pairs in the RNA exit channel does not place RNAP in a persistent hyper-translocated state. Furthermore, these findings show that TECprobe-LM can be used to evaluate interactions between RNAP and nascent RNA that facilitate RNA folding.The basic TECprobe-LM procedure is designed to assess one cotranscriptional RNA folding event at a time. Nonetheless, it was possible to assess two sequential pfl ZTP aptamer folding events by adjusting NTP concentration and depletion conditions so that RNAP transcribed forward several nucleotides upon release of the NPOM cage. This approach took advantage of the observation that RNAP was prone to transcribing 1-2 nucleotides forward in this sequence context even after extensive washing to deplete NTPs and may not be easily generalizable. The simplest way to implement a TECprobe-LM experiment for >2 folding intermediates is to position RNAP at an initial NPOM-caged-dT modification and subsequently walk RNAP to specific downstream sites by resuming transcription in the absence of one or more NTPs. However, the complexity of a TECprobe-LM experiment increases substantially as more folding intermediates are assessed both because the efficiency of walking RNAP is sequence-dependent and because TECs that do not resume transcription after an initial walk may resume transcription during a subsequent walk. Furthermore, every additional intermediate that is assessed in a TECprobe-LM experiment requires two or four additional samples depending on whether a pre-wash control sample is collected. For these reasons, we suggest that the most straightforward way to assess the validity of a sequence of RNA folding transitions is to perform several independent TECprobe-LM experiments that assess overlapping folding transitions as we have done in the ZTP and fluoride riboswitch systems.In addition to directly assessing cotranscriptional RNA folding events, TECprobe-LM establishes an experimental framework that could potentially be used to assess the kinetic processes that contribute to cotranscriptional RNA structure formation. For example, the effect of transcription rate on a cotranscriptional RNA folding transition could be assessed by varying nucleotide concentration when RNAP is chased from the NPOM-caged-dT site to the biotin-streptavidin roadblock or by using RNAP variants that are more or less prone to transcription pausing53,54,55,56. When assessing folding transitions that occur on the order of seconds, it may be possible to probe folding intermediates using a quench-flow instrument, as has been done previously to measure B. subtilis RNase P tertiary structure formation under equilibrium folding conditions57. For this application, it would be necessary to identify an NPOM-caged-dT stall site from which E. coli RNAP resumes transcription synchronously or to include the transcription elongation factor GreB58, which resolves backtracked TECs that would otherwise remain arrested when RNAP is chased downstream59,60,61. Furthermore, chemical probes that react with RNA rapidly, such as benzoyl cyanide57,62 (t1/2 = 250 ms, used in this work), nicotinoyl azide63,64 (ps timescale), and hydroxyl radical footprinting65,66,67,68 (ms timescale) will be necessary to capture changes in RNA structure that occur over short intervals. Nicotinoyl azide, which is used in the Light-Activated Structural Examination of RNA (LASER) strategy for RNA chemical probing63,64, could be of particular interest for time-resolved cotranscriptional RNA structure probing experiments because it is activated by 310 nm UV light, which eliminates the need to mix the probe with the sample at each time point. The TECprobe-LM framework could also facilitate atomic mutagenesis experiments in which modified nucleotides are site-specifically incorporated into nascent RNA downstream of the photoreversible transcription roadblock prior to chasing RNAP downstream to the terminal biotin-streptavidin roadblock. In these ways, TECprobe-LM can enable the observation of cotranscriptional RNA folding directly by chemical probing.MethodsAn inventory of all reagents used in this study, including the manufacturer and catalog number of each reagent, is provided in Supplementary Data 1.OligonucleotidesAll oligonucleotides were purchased from Integrated DNA Technologies. A detailed description of all oligonucleotides including sequence, modifications, and purifications is presented in Supplementary Table 1.ProteinsQ5 High-Fidelity DNA Polymerase, Vent (exo-) DNA polymerase, Sulfolobus DNA Polymerase IV, Lambda Exonuclease, E. coli RNA Polymerase holoenzyme, Mth RNA Ligase (as part of the 5’ DNA Adenylation kit), T4 RNA Ligase 2 truncated KQ, ET SSB, RNase H, and RNase If were purchased from New England Biolabs. TURBO DNase, SuperaseIN, SuperScript II, and BSA were purchased from ThermoFisher. Streptavidin was purchased from Promega.DNA template purificationDNA templates that contained an internal NPOM-caged-dT and 5’ biotin modification were prepared under 592 nm amber light by one of two strategies, which are described in detail below. In the first strategy, PCR amplification was performed using a primer that contained the NPOM-caged-dT and 5’ biotin modifications. Translesion DNA synthesis was then performed to fill in the 5’ overhang that results from the NPOM-caged-dT modification blocking complete synthesis of the non-transcribed DNA strand. We previously used this strategy to synthesize DNA templates that contain internal biotin-TEG69, desthiobiotin-TEG, etheno-dA, and amino linker modifications43, however the success of translesion synthesis past the NPOM-caged-dT modification was variable depending on DNA template sequence. In the second strategy, we circumvented this issue by performing the initial PCR amplification using a 5’-phosphorylated reverse primer so that the transcribed DNA strand could be selectively degraded by lambda exonuclease and resynthesized using a primer that contained the NPOM-caged-dT and 5’ biotin modifications.In all DNA template preparations, an unmodified linear dsDNA template was first amplified from plasmid DNA. PCR was performed as three 100 μl reactions containing 1X Q5 Reaction Buffer, 200 μM dNTPs, 250 nM PRA1_NoMod.F (Supplementary Table 1), 250 nM of a reverse primer that varied depending on the target sequence and the DNA template preparation strategy (Supplementary Tables 1 and 2), 0.2 ng/μl plasmid DNA, and 0.02 U/μl Q5 DNA polymerase using the following thermal cycler protocol with a heated lid set to 105 °C: 98 °C for 30 s, [98 °C for 10 s, 65 °C for 30 s, 72 °C for 20 s] x 30 cycles, 72 °C for 2 min, hold at 12 °C. For DNA template preparations in which the NPOM-caged-dT was added by PCR and translesion DNA synthesis, the 5’ end of the reverse primer was either positioned 1 nt before the position at which the NPOM-caged-dT modification would be located in the final dsDNA template or the primer overlapped this position. For DNA template preparations in which the NPOM-caged-dT modification was added by lambda exonuclease treatment and primer extension, the reverse primer matched the sequence of the NPOM-caged-dT-modified reverse primer that would be used for primer extension, but did not contain the NPOM-caged-dT modification and was 5’-phosphorylated. Linear dsDNA was then ethanol precipitated, purified by UV-free agarose gel extraction using a QIAquick gel extraction kit, and quantified using the Qubit dsDNA Broad Range Assay kit with a Qubit 4 Fluorometer69.When preparing DNA templates by PCR and translesion DNA synthesis, the NPOM-caged-dT modification was incorporated in a second 8 × 100 μl PCR containing 1X Q5 Reaction Buffer, 1X Q5 High GC Enhancer, 200 μM dNTPs, 250 nM PRA1_NoMod.F (Supplementary Table 1), 250 nM of an NPOM-caged-dT modified reverse primer that varied depending on target sequence (Supplementary Tables 1 and 2), 20 pM linear dsDNA template prepared as described above, and 0.02 U/μl Q5 DNA polymerase using the following thermal cycler protocol with a heated lid set to 105 °C: 98 °C for 30 s, [98 °C for 10 s, 65 °C for 30 s, 72 °C for 20 s] x 35 cycles, 72 °C for 2 min, hold at 12 °C. PCRs were then purified using a QIAquick PCR purification kit according to the manufacturer’s protocol and translesion DNA synthesis was performed by incubating three 100 μl reactions that contained 1X ThermoPol Buffer, 200 μM dNTPs, 0.02 U/μl Sulfolobus DNA polymerase IV, 0.02 U/μl Vent (exo-) DNA polymerase, and the purified PCR products at 55 °C for one hour. The resulting dsDNA templates were purified a second time using a QIAquick PCR purification kit according to the manufacturer’s protocol, eluted into 25 μl of 10 mM Tris (pH 8.0) per translesion synthesis reaction and quantified using the Qubit dsDNA Broad Range Assay kit with a Qubit 4 Fluorometer.When preparing DNA templates by PCR, lambda exonuclease treatment, and primer extension, lambda exonuclease treatment was performed as 50 μl reactions containing 1X Lambda Exonuclease Reaction Buffer, up to 600 nM linear dsDNA template prepared as described above, and 0.1 U/μl lambda exonuclease. Reactions were incubated at 37 °C for 30 min, stopped by adding 1 μl of 500 mM EDTA (pH 8.0), and incubated at 75 °C for 10 min to heat inactivate lambda exonuclease. The sample volume was raised to 150 μl by adding 100 μl of 10 mM Tris (pH 8.0), mixed with an equal volume (150 μl) of phenol:chloroform:isoamyl alcohol (25:24:1) by vortexing, centrifuged at 18,000 x g and 4 °C for 5 min, and the aqueous supernatant was collected into a new tube. The samples were then ethanol precipitated by adding 0.1 volumes (15 μl) of 3 M sodium acetate (pH 5.5), 3 volumes (450 μl) of 100% ethanol, and 1.5 μl of Glycoblue coprecipitant, and chilling at −20 °C overnight or −70 °C for 30–60 min. The samples were centrifuged at 18,000 x g and 4 °C for 30 min, the supernatant was discarded, and the samples were washed by adding 1 ml of 70% ethanol and inverting the tube gently. The samples were centrifuged at 18,000 x g and 4 °C for 5 min and the supernatant was discarded. The samples were briefly spun in a mini centrifuge, and residual liquid was discarded. Each pellet was then resuspended in 50 μl of 10 mM Tris (pH 8.0). The NPOM-caged-dT modification was then incorporated by converting the purified ssDNA into dsDNA in three 100 μl primer extension reactions containing 1X ThermoPol Buffer, 200 μM dNTPs, 300 μM NPOM-caged-dT modified primer (Supplementary Tables 1 and 2), ssDNA that was generated and purified as described above, and 0.02 U/μl Vent (exo-) DNA polymerase using the thermal cycler program: 95 °C for 3 min, 65 °C for 10 min, 72 °C for 10 min, hold at 12 °C. 0.5 μl of thermolabile exonuclease I was added to each sample and the samples were incubated at 37 °C for 4 min and placed on ice. Thermolabile exonuclease I was then heat-inactivated by incubating the samples on a thermal cycler block that had been pre-heated to 80 °C for 1 min. The primer extension reactions were purified using a QIAquick PCR purification kit according to the manufacturer’s protocol, eluted into 25 μl of 10 mM Tris (pH 8.0) per primer extension reaction and quantified using the Qubit dsDNA Broad Range Assay kit with a Qubit 4 Fluorometer. The sequences of all DNA templates are provided in Supplementary Table 3.When preparing randomly biotinylated DNA templates for TECprobe-VL experiments, the initial amplification from plasmid DNA was performed as described above except that the reverse primer HP4_5bio.R, which contains a 5’ biotin modification, was used. Randomly biotinylated DNA templates were then PCR amplified from the 5’ biotinylated linear DNA template using Vent (exo-) DNA polymerase and primers PRA1_NoMod.F and HP4_5bio.R20. 200 μl PCRs contained 1X ThermoPol Buffer, 250 nM PRA1_NoMod.F, 250 nM HP4_5bio.R, 20 pM template DNA, 0.02 U/μl Vent (exo-) DNA polymerase, 200 μM dNTP Solution Mix, and a concentration of biotin-11-dNTPs (PerkinElmer, Biotium) that favored the incorporation of ~2 biotin modifications13 in the transcribed region of each DNA template. PCR was performed as two 100 μl reactions in thin-walled tubes using the following thermal cycler protocol with a heated lid set to 105 °C: 95 °C for 3 min, [95 °C for 20 s, 58 °C for 30 s, 72 °C for 30 s] x 30 cycles, 72 °C for 5 min, hold at 12 °C. PCR products were purified as described below in the section SPRI bead purification of DNA, eluted into 50 μl of 10 mM Tris-HCl (pH 8.0), and quantified using the Qubit dsDNA Broad Range Assay Kit (Invitrogen) with a Qubit 4 Fluorometer (Invitrogen). A step-by-step protocol for this procedure is available70.SPRI bead purification of DNASPRI beads were prepared in-house using the ‘DNA Buffer’ variation of the procedure by Jolivet and Foley71, and DNA was purified according to an established procedure20. Briefly, samples were mixed with an equal volume of SPRI beads, incubated at room temperature for 5 min, and placed on a magnetic stand for 3 min so that the beads collected on the tube wall. The supernatant was aspirated and discarded, and the beads were washed twice by adding a volume of 70% ethanol at least 200 μl greater than the combined volume of the sample and SPRI beads to the tube without disturbing the bead pellet while it remained on the magnetic stand. The samples were incubated at room temperature for 1 min before aspirating and discarding the supernatant. Residual ethanol was evaporated by placing the open microcentrifuge tube in a 37 °C dry bath for ~15 s with care taken to ensure that the beads did not dry out. Purified DNA templates were eluted by resuspending the beads in a variable amount of 10 mM Tris-HCl (pH 8.0) (depending on the procedure, details are in each relevant section), allowing the samples to sit undisturbed for 3 min, placing the sample on a magnetic stand for 1 min so that the beads collected on the tube wall, and transferring the supernatant, which contained purified DNA, into a screw-cap tube with an O-ring.Denaturing PAGE analysis of reversible transcription roadblockingSingle-round in vitro transcription was performed as described below for TECprobe-LM except that the volume of the transcription reaction was 87.5 μl and all wash and resuspension volumes were scaled accordingly. 25 μl aliquots of the transcription reaction were removed, transferred to 75 μl of TRIzol LS and vortexed at three points: First, the pre-wash sample was removed after RNAP had arrested at the NPOM-caged-dT roadblock. Second, the pre-chase sample was removed after nucleotides were depleted by washing the beads. Third, the post-chase sample was removed after RNAP was chased to the terminal biotin-streptavidin roadblock. 20 μl of chloroform was added to each sample and the samples were mixed by vortexing and inversion. The samples were centrifuged at 18,000 x g for 5 min and the aqueous phase was collected, mixed with 1.2 μl of GlycoBlue Coprecipitant and 50 μl of ice-cold isopropanol, incubated at room temperature for 15 min, and centrifuged at 18,000 x g for 15 min. The supernatant was discarded and the resulting pellet was washed with 200 μl of ice cold 70% ethanol by gently inverting the tube. The samples were centrifuged at 18,000 x g for 2 min and the supernatant was discarded. The samples were briefly spun in a mini centrifuge to pull down residual liquid, and the residual liquid was discarded. The samples were resuspended in 25 μl of 1X Turbo DNase Buffer, mixed with 0.75 μl of Turbo DNase, and incubated at 37 °C for 15 min. Each sample was mixed with 75 μl of TRIzol LS and purified by TRIzol extraction and isopropanol precipitation as described above. The resulting pellets were resuspended in 15 μl of formamide loading dye (90% v/v deionized formamide, 1X Transcription Buffer (defined below), 0.05% w/v bromophenol blue), denatured by incubating at 95 °C for 5 min and fractionated using an 8% polyacrylamide gel prepared using the SequaGel 19:1 Denaturing Gel system (National Diagnostics) for a Mini-PROTEAN Tetra Vertical Electrophoresis Cell. Denaturing conditions were achieved by filling the outer buffer chamber so that buffer covered only ~1 cm of the gel plates, pre-running the gel at 480 V for 30 min, and running the gel at 480 V for ~10 min69. Gels were stained with 1X SYBR Gold Nucleic Acid Stain in 1X TBE for 10 min and scanned on a Typhoon RGB Biomolecular Imager.TECprobe-LMAll steps of the TECprobe-LM procedure were performed under 592 nm amber light until the samples were irradiated with 365 nm UV light. One sample volume is defined as 25 μl.5 μl of 10 mg/ml Dynabeads MyOne Streptavidin C1 beads (Invitrogen) per sample volume were equilibrated in Buffer TX (1X Transcription Buffer, 0.1% (v/v) Triton X-100)69. Briefly, after placing the beads on a magnetic stand and removing the storage buffer, the beads were resuspended in 500 μl of Hydrolysis Buffer (100 mM NaOH, 50 mM NaCl) and incubated at room temperature for 10 min with rotation. Hydrolysis Buffer was removed, and the beads were resuspended in 1 ml of High Salt Wash Buffer (50 mM Tris-HCl (pH 7.5), 2 M NaCl, 0.5% (v/v) Triton X-100), transferred to a new tube, and washed by rotating for 5 min at room temperature. High Salt Wash Buffer was removed, and the beads were resuspended in 1 ml of Binding Buffer (10 mM Tris-HCl (pH 7.5), 300 mM NaCl, 0.1% (v/v) Triton X-100), transferred to a new tube, and washed by rotating for 5 min at room temperature. After removing Binding Buffer, the beads were washed twice with 500 μl of Buffer TX by resuspending the beads, transferring them to a new tube, washing with rotation for 5 min at room temperature, and removing the supernatant. After washing the second time with Buffer TX, the beads were resuspended to a concentration of ~2 μg/μl in Buffer TX (25 μl per sample volume) and stored on ice until use.For each set of TECprobe-LM samples (including pre-wash/pre-UV, pre-chase, and post-chase samples) a 165 μl single-round in vitro transcription reaction (6.6 sample volumes) containing 1X Transcription Buffer (20 mM Tris-HCl (pH 8.0), 50 mM KCl, 1 mM DTT, 0.1 mM EDTA (pH 8.0)), 0.1 mg/ml BSA, 0.05% Tween-20, 10 nM DNA template, and 0.024 U/μl E. coli RNAP holoenzyme was prepared on ice and incubated at 37 °C for 15 min to form open promoter complexes. A 150 μl aliquot of 2 μg/μl streptavidin-coated magnetic beads that were equilibrated as described above were placed on a magnetic stand and the supernatant was removed. The beads were then resuspended using the transcription reaction and incubated at room-temperature with end-over-end rotation at 15 rpm for 15 min to immobilize the template DNA. The sample was then briefly spun in a mini centrifuge and placed onto a magnetic stand for 1 min to collect the beads on the tube wall. The supernatant was removed and the beads were resuspended in 660 μl of Wash Buffer 1 (1X Transcription Buffer, 0.05% Tween-20, 0.1 mg/ml BSA), returned to the magnetic stand, and the supernatant was removed. The beads were then washed using 660 μl of Wash Buffer 1 a second time. The beads were resuspended in 22.5 μl of Reaction Buffer 1 per sample volume and incubated at 37 °C for 2 min before transcription was initiated by adding 16.5 μl of 10X Start Solution (100 mM MgCl2 and 0.1 mg/ml rifampicin). Upon addition of 10X start solution, the transcription reaction contained 1X Transcription Buffer, 0.1 mg/ml BSA, 0.05% Tween-20, 50 μM or 100 μM NTPs, 10 mM MgCl2, and 10 μg/μl rifampicin. In ZTP riboswitch experiments, the transcription reaction also contained 1 mM ZMP and 2% DMSO (1 mM ZMP samples) or 2% DMSO (0 mM ZMP samples). In fluoride riboswitch experiments, the transcription reaction also contained 10 mM NaF when fluoride was present. The transcription reaction was incubated at 37 °C for 2 min to allow RNAP to transcribe to the NPOM-caged-dT modification.In all experiments except that of Fig. 3, the pre-wash sample was then chemically probed by mixing 25 μl of the transcription reaction with 400 mM benzoyl cyanide57,62 (BzCN, modified channel) that was dissolved in anhydrous DMSO so that the chemical probing reaction contained 40 mM BzCN and 10% DMSO. In the untreated control sample, 25 μl of the transcription reaction was mixed with anhydrous DMSO (untreated channel) so that the sample contained 10% DMSO. Each channel of the pre-wash sample was then mixed with 75 μl of TRIzol LS and vortexed thoroughly. The remaining transcription reaction was placed on a magnetic stand, the supernatant was discarded, and the beads were resuspended in 840 μl of Wash Buffer 2. For all samples, Wash Buffer 2 contained 1X Transcription Buffer, 0.1 mg/ml BSA, 0.05% Tween-20, 10 mM MgCl2, and 10 μg/ml rifampicin. For ZTP riboswitch experiments performed in the presence of ZMP, Wash Buffer 2 also contained 0.1 mM ZMP and 0.2% DMSO. For ZTP riboswitch experiments performed in the absence of ZMP, Wash Buffer 2 also contained 0.2% DMSO. For fluoride riboswitch experiments performed in the presence of fluoride, Wash Buffer 2 also contained 1 mM NaF. The beads were transferred to a new tube, placed on a magnetic stand, and the supernatant was removed. The beads were resuspended in 840 μl of Wash Buffer 2 a second time, transferred to a new tube, placed on a magnetic stand, and the supernatant was removed. The beads were then resuspended in 840 μl of Wash Buffer 2 a third time, transferred to a new tube, incubated on an end-over-end rotator at room temperature for 5 min, briefly spun down in a mini centrifuge, placed on a magnetic stand, and the supernatant was removed. The beads were then resuspended in 24.7 μl of Reaction Buffer 2 per sample volume which, upon completing the reaction by adding NTPs in a later step, contained 1X Transcription Buffer, 0.1 mg/ml BSA, 0.05% Tween-20, 10 mM MgCl2, 10 μg/ml rifampicin. For ZTP riboswitch experiments, Reaction Buffer 2 also contained 1 mM ZMP and 2% DMSO (1 mM ZMP samples) or 2% DMSO (0 mM ZMP samples). For fluoride riboswitch experiments, Reaction Buffer 2 also contained 10 mM NaF when fluoride was present.In the experiment shown in Fig. 3 in which three-point chemical probing was performed, samples were washed immediately after RNAP had transcribed to the NPOM-caged-dT roadblock as follows: The sample was placed on a magnetic stand and the supernatant was discarded. The beads were resuspended in 840 μl of Wash Buffer 2, transferred to a new tube, incubated on an end-over-end rotator at room temperature for 5 min, briefly spun down in a mini centrifuge, placed on a magnetic stand, and the supernatant was removed. The beads were then resuspended in 840 μl of Wash Buffer 2 and washed as described above a second time. For experiments performed in the presence of ZMP, Wash Buffer 2 was supplemented with 0.1 mM ZMP and 0.2% DMSO. For experiments performed in the absence of ZMP, Wash Buffer 2 was supplemented with 0.2% DMSO. The beads were then resuspended in 24.7 μl of Reaction Buffer 2 per sample volume and incubated at 37 °C for 2 min. The pre-UV sample was then chemically probed as described above.The sample was placed on a custom-built microcentrifuge tube irradiator69 and irradiated with 10 mW/cm2 365 nm UV light for 3 min, with mixing after every minute. Irradiation can also be performed using a 365 nm UV LED flashlight (e.g. Waveform Lighting, cat. # 7023), as demonstrated for a photocleavable 2-nitrobenzyl linker (PC-linker) modification by Zou and colleagues72. The intensity of the UV light source can be assessed using a UVA/B digital light meter calibrated to 365 nm (e.g. General, cat. #UV513AB). The sample was incubated at 37 °C for 2 min and the pre-chase sample was chemically probed and stopped with TRIzol LS as described above for the pre-wash sample. 1.3 μl of 5 mM NTPs were added to the remaining sample and the reaction was incubated at 37 °C for 2 min to allow RNAP to transcribe to the terminal biotin-streptavidin roadblock. The post-chase sample was then chemically probed and stopped with TRIzol LS as described above for the pre-wash sample.The samples were then TRIzol extracted and converted to cDNA and indexed dsDNA Illumina libraries by ligating an adapter to the RNA 3’ end, performing solid-phase error-prone reverse transcription, and degrading the RNA using the TECprobe-VL protocol for sequencing library preparation20. The original procedure is presented below in the section Sequencing library preparation for TECprobe-LM and TECprobe-VL experiments.TECprobe-VLSingle-round in vitro transcription reactions for TECprobe-VL experiments were performed as 60 μl reactions containing 1X Transcription Buffer, 0.1 mg/ml Molecular Biology-Grade BSA, 100 μM high-purity NTPs, 10 nM randomly biotinylated template DNA, and 0.024 U/μl E. coli RNA polymerase holoenzyme. At the time of preparation, each TECprobe-VL reaction was 48 μl due to the omission of 10X (1 μM) streptavidin (Promega) and 10X Start Solution from the reaction. Single-round in vitro transcription reactions were incubated at 37 °C for 10 min to form open promoter complexes. 6 μl of 1 μM streptavidin was then added for a final concentration of 100 nM streptavidin, and reactions were incubated for an additional 10 min at 37 °C. Transcription was initiated by adding 6 μl of 10X Start Solution to the reaction for a final concentration of 10 mM MgCl2 and 10 μg/ml rifampicin. The transcription reaction was incubated at 37 °C for 2 min before the sample was split into 25 μl aliquots and mixed with 2.8 μl of 400 mM BzCN dissolved in anhydrous DMSO [(+) sample)] or with anhydrous DMSO [(-) sample]57,62. The samples were then mixed with 75 μl of TRIzol LS reagent (Invitrogen) and vortexed to stop the in vitro transcription reaction. Sequencing libraries were then prepared as described below.Sequencing library preparation for TECprobe-LM and TECprobe-VL experimentsSequencing library preparation was performed exactly as described previously. The original procedure20 is presented below.RNA purificationSamples, which contained 27.8 μl of the cotranscriptional RNA chemical probing reaction in 75 μl of TRIzol LS, were extracted as follows: 20 μl of chloroform was added to each sample, and the samples were mixed by vortexing and inverting the tube and centrifuged at 18,500 x g and 4 °C for 5 min. The aqueous phase was transferred to a new tube and precipitated by adding 1.5 μl of GlycoBlue Coprecipitant (Invitrogen) and 50 μl of ice-cold isopropanol and incubating at room temperature for 15 min. The samples were centrifuged at 18,500 x g and 4 °C for 15 min, the supernatant was aspirated and discarded, 500 μl of ice-cold 70% ethanol was added to each sample, and the tubes were gently inverted to wash the samples. The samples were centrifuged at 18,500 x g and 4 °C for 2 min and the supernatant was aspirated and discarded. The samples were centrifuged again briefly to pull down residual liquid, which was aspirated and discarded. The pellet was then resuspended in 25 μl of 1X TURBO DNase buffer (Invitrogen), mixed with 0.75 μl of TURBO DNase (Invitrogen), and incubated at 37 °C for 15 min. 75 μl of TRIzol LS reagent was added to stop the reactions and a second TRIzol extraction was performed as described above, except that the pellet was resuspended in 5 μl of 10% (v/v) DMSO.RNA 3’ adapter ligation9N_VRA3 adapter oligonucleotide (Supplementary Table 1) was pre-adenylated with the 5’ DNA Adenylation Kit (New England Biolabs) according to the manufacturer’s protocol at a 5X scale. Briefly, 100 μl of a master mix that contained 1X DNA Adenylation Buffer (New England Biolabs), 100 μM ATP, 5 μM 9N_VRA3 oligo, and 5 μM Mth RNA Ligase (New England Biolabs) was split into two 50 μl aliquots in thin-walled PCR tubes and incubated at 65 °C in a thermal cycler with a heated lid set to 105 °C for 1 h. Following the reaction, 150 μl of TRIzol LS reagent was added to each 50 μl reaction and the samples were extracted as described above in the section RNA purification, except that reaction volumes were scaled to account for the 50 μl reaction volume (40 μl of chloroform was added to the sample-TRIzol mixture and 100 μl of isopropanol was added during the precipitation step). Samples were pooled by resuspending the pellets from each TRIzol extraction in a single 25 μl volume of TE Buffer (10 mM Tris-HCl (pH 7.5), 0.1 mM EDTA). The concentration of the adenylated oligonucleotide was determined using the Qubit ssDNA Assay Kit (Invitrogen) with a Qubit 4 Fluorometer. The molarity of the linker was calculated using 11,142 g/mol as the molecular weight. The adenylation reaction was assumed to be 100% efficient. The linker was diluted to 0.9 μM and aliquoted for future use; aliquots were used within 3 freeze-thaw cycles.20 μl RNA 3’ adapter ligation reactions were performed by combining purified RNA in 5 μl of 10% DMSO (v/v) from the RNA purification section with 15 μl of an RNA ligation master mix such that the final 20 μl reaction contained purified RNA, 2.5% (v/v) DMSO, 1X T4 RNA Ligase Buffer (New England Biolabs), 0.5 U/μl SuperaseIN (Invitrogen), 15% (w/v) PEG 8000, 45 nM 5’-adenylated 9N_VRA3 adapter, and 5 U/μl T4 RNA Ligase 2, truncated, KQ (New England Biolabs). The samples were mixed by pipetting and incubated at 25 °C for 2 h.SPRI bead purification of RNAExcess 9N_VRA3 3’ adapter oligonucleotide was depleted using a modified SPRI bead purification that contains isopropanol73. 17.5 μl of nuclease-free water and 40 μl of freshly aliquoted anhydrous isopropanol (Sigma-Aldrich) were added to each 20 μl RNA ligation reaction, and the samples were mixed by vortexing. Each sample was then mixed with 22.5 μl of SPRI beads so that the concentration of PEG 8000 was 7.5% (w/v) and the concentration of isopropanol was 40% (v/v) in a sample volume of 100 μl. The samples were incubated at room temperature for 5 min, and placed on a magnetic stand for at least 3 min so that the beads collected on the tube wall. The supernatant was aspirated and discarded, and the beads were washed twice by adding 200 μl of 80% (v/v) ethanol to the tubes without disturbing the bead pellet while it remained on the magnetic stand, incubating the samples at room temperature for 1 min, and aspirating and discarding the supernatant. After discarding the second 80% (v/v) ethanol wash, the sample was briefly spun in a mini centrifuge, and placed back onto a magnetic stand for 1 min to collect the beads on the tube wall. The supernatant was aspirated and discarded, and the beads were briefly (100 or