Transcriptional suppression is PQS position-dependent
Next, we asked if the PQS-NT positions impact transcription. We prepared linearized and enclosed plasmid with PQS inserted at 10, 30, and 60 bp downstream of TSS in the non-template strand (NT) (NT-sp10, NT-sp30, NT-sp60) (Fig. 1d). Linearized DNAs with NT-sp10, -sp30, and -sp60 showed no significant difference in RNA production rate, indicating that PQS position does not affect transcription in linear DNA (Fig. 1d). However, the same set of PQS in plasmid exhibited notable differences in RNA production rate. In particular, NT-sp10 and NT-sp60 showed ~ 40% and ~ 20% reduced RNA production rate than NT-sp30, respectively. Further tests on the PQS-NT positions at 10, 15, 20, 25, 30, 35, 45, and 60 bp downstream of TSS (Fig. 1e) displayed an interesting pattern i.e the PQS located proximal to TSS (NS-sp10, 15, 20) displayed the highest level of transcription suppression, which is gradually alleviated downstream (NT-sp25, 30) and again suppressed further downstream (NT-sp35, 45, 60). Importantly, all constructs led to a lower rate of RNA production than the control. Taken together, PQS in supercoiled DNA suppresses transcription position-dependent, while the same PQS in linear DNA enhances transcription. Such contrast emphasizes the unique regulatory role of PQS in supercoiled DNA.
PQS-driven DNA topological changes mediate transcription suppression through G4-stabilized R-loop formation.
Since the PQS-mediated transcription suppression was exclusive to supercoiled DNA, we hypothesized that DNA undergoes PQS-driven topological changes during transcription. Two PQS positions, NT-sp10 and NT-sp30, were chosen based on the high and low levels of suppression, respectively (Fig. 1d and e). We performed in vitro transcription reactions with NT-sp10 and NT-sp30 in linear and supercoiled DNA and subsequently applied the transcribed samples on a 1% agarose gel. In supercoiled DNA, both NT-sp10 and NT-sp30 underwent a substantial topological shift from primarily supercoiled to a relaxed or nicked state during transcription (Fig. 2b, left panel). In contrast, their linear counterparts remained in the same position (Fig. 2b, right panel). The cross-section analysis of the gel image reveals that the upward band shift is more pronounced in NT-sp10 than in NT-sp30 (Fig. 2c). Based on the higher suppression seen in NT-sp10, the shifted bands likely correspond to DNA conformations that suppress transcription.
Next, we tested if the observed band shift is due to R-loop formation, as PQS-NT is prone to lead to R-loop formation39 and R-loops alleviate torsional stress in DNA molecules.48 When we treated the transcribed samples with RNase H, which explicitly degrades RNA in the R-loop, all the shifted bands from both NT-sp10 and NT-sp30 returned to the original, i.e., primarily negatively supercoiled state of the plasmid (Fig. 2b and c, lanes 1 and 3). The electrophoretic mobility shift assay with S9.6, a monoclonal antibody specific to R-loops, further confirmed that all the shifted bands consisted of R-loops (Supplementary Fig. 1). These results indicate that (i) R-loop formation is responsible for the observed band shift for both NT-sp10 and NT-sp30, (ii) the NT-sp10 induces a higher level of R-loop than in NT-sp30, and (iii) a higher R-loop in NT-10 drive stronger transcriptional repression. The R-loop is an inhibitory structure for transcription under supercoiled DNA conditions. This relationship is demonstrated by the gradient of R-loop forming propensity: highest to lowest R-loops are detected for NT-sp10, -sp15, -sp20, -sp30, -sp35 (Fig. 2d) which is inversely related to the transcription rate (Fig. 1e).
Furthermore, we measured the kinetics of R-loop formation in NT-sp10 and NT-sp30 by taking samples at 10, 20, 30, and 60 minutes of transcription reaction and analyzing them by the gel shift assay (Supplementary Fig. 2). While NT-sp10 displayed a rapid topological shift within the first 5 minutes, NT-sp30 showed a more gradual shift over 60 minutes (Fig. 2e). The faster shift seen in NT-sp10 than in NT-sp30 indicates that the early-stage R-loop formation leads to a more drastic reduction in transcription in NT-sp10 (Fig. 2f). To test if G4 contributes to the R-loop mediated transcriptional suppression, we applied G4 destabilizing (no mono-valent ion, LiCl) and stabilizing (KCl) conditions (Supplementary Fig. 3).49,50 The G4 stabilizing condition led to increased topological relaxation only for NT-sp30 but not for control or T-sp30, suggesting that the formation of G4 structures in NT-sp30 promotes topological shifts, i.e., R-loop formation. These findings further support that G4 forms and plays a pivotal role in stabilizing R-loop structures.
Real-time single-molecule detection of co-transcriptional R-loop and G4 formation
We employed single-molecule assays to probe the real-time dynamics underlying DNA conformational changes and topological shifts during transcription. We employed single-molecule (sm) FRET to measure DNA structural change.51,52 and the single-molecule protein-induced fluorescence enhancement (smPIFE) to measure RNAP binding.53,54 We constructed linearized, relaxed, and negatively supercoiled 3.7 kb plasmid DNAs with two types of labeling strategies: [FRET1] for monitoring G4/R-loop formation and RNAP movement and [FRET2] for monitoring the transcription initiation process (Fig. 4a). Using a nicking enzyme-based internal labeling method,55 Cy3, Cy5, and biotin-labeled oligomers were annealed to the 3.7 kb-sized vector and sealed through ligation (Fig. 4b). Negatively supercoiled DNA was prepared by exposing the annealed DNA to ethidium bromide during the ligation process (Fig. 4b left, lane 1). Without ethidium bromide, it generates relaxed DNA (Fig. 4b left, lane 2). As shown in Fig. 4b left, we imaged the DNA constructs on 1% agarose gel without post-staining because the constructs were already labeled with Cy3 and Cy5 fluorophores. By post-staining the gel, we confirmed that the Cy3 and Cy5 labeled supercoiled DNA exhibited a similar level of superhelicity with the plasmid purified from E. coli (Fig. 4b, right). We verified that the Cy3 and Cy5 incorporated into 3.7 kb plasmid were successfully immobilized on the surface and showed the expected FRET values (E) of ~ 0.3 for both constructs (Fig. 4c). The linearized DNA was prepared by cutting the relaxed DNA with a single-cut restriction enzyme (Supplementary Fig. 5).
Transient R-loops promote G4/R-loop formation.
[FRET1] has Cy3 and Cy5 across NT-sp30, labeled at 27 bp (+ 27) and 47 bp (+ 47) downstream of TSS, respectively. Based on our previous study, the expected FRET values for the DNA-only, DNA with R-loop, and DNA with G4 are ~ 0.3, ~ 0.7, and ~ 0.9, respectively.39 Additionally, we expect to visualize RNAP elongation near Cy3 at + 27 via the PIFE signal (Fig. 5a).
For the relaxed [FRET1] construct, we observed primarily short-lived PIFE peaks without FRET change, which reports on successive RNAP elongation without any structural changes in the DNA (Fig. 5b).56–58 The short-lived PIFE peaks indicate elongating RNAP as the peak frequency increased as a function of RNAP concentration (Supplementary Fig. 6). However, no G4 and R-loop formations were observed in the relaxed construct within the first 150 seconds of observation time. In contrast, the supercoiled [FRET1] construct displayed two distinct patterns of FRET and PIFE signals: First, short-lived and frequent PIFE-FRET (E ~ 0.7) tandem peaks appeared (Fig. 5c and d). The time delay between PIFE and FRET peaks indicates that the PQS-bearing DNA segment undergoes a brief R-loop formation when RNAP transcribes through the PQS region. Second, after multiple rounds of short-lived PIFE-FRET transitions, the mid-FRET (E ~ 0.7) converts to a high-FRET state (E ~ 0.9) (Fig. 5e), signifying the transient R-loop to G4 transition respectively (Fig. 5c and e). Once it transitions to high-FRET, the state remains stable. The FRET histogram analysis revealed that the high FRET peak diminished in G4 destabilizing conditions (Fig. 5f, no M+), confirming the long-lived high-FRET as a stably folded G4 state (Fig. 5f). The RNase H treatment led to the disappearance of the mid-FRET (~ 0.7) peak, as expected from the R-loop removal. The 0.7 FRET peak is not distinct because the R-looped state is highly transient in supercoiled DNA, as shown in Fig. 5c-e. Remarkably, 97.5% (119/122) of molecules that folded into G4 (E ~ 0.9) exhibited stepwise FRET transition from E ~ 0.7 to E ~ 0.9 (See more representative traces in Supplementary Fig. 7). This observation implies that the transiently formed R-loop is required to form a stable G4 structure. The transiently formed R-loop likely creates a locally unwound region, which may reduce the energetic barrier for PQS to fold into G4. In addition, supercoiled [FRET1] didn’t show a long-lived R-looped state, unlike linear [FRET1] (Supplementary Fig. 8). The gel retardation assay showed that G4/R-loop induces topological relaxation, which contributes to transcriptional suppression (Fig. 2 and Supplementary Fig. 2–3). Based on the result, the stable high FRET state likely encompasses both R-loop and G4 (G4/R-loop).
Supercoiling accelerates RNAP loading, which stimulates G4/R-loop formation.
One striking difference is that the supercoiled DNA construct exhibited a significantly higher frequency of PIFE spikes than the relaxed construct before forming the stable high-FRET state (E ~ 0.9) (Fig. 5d). When we quantified the number of PIFE peaks (Fig. 5g), supercoiled DNA showed an approximately ten-fold increase in transcription rate compared to the linearized and relaxed DNA. We postulated that the increased transcription rate that involves frequent RNAP loading would lead to the underwinding of the upstream DNA, which may increase the probability of G4/R-loop formation. As expected, supercoiled DNA exhibited a significantly higher fraction of G4/R-loop (high FRET) than relaxed DNA, even with a 5-fold higher RNAP concentration (Fig. 5h and Supplementary Fig. 8). This result is consistent with the shifted gel bands corresponding to the R-loop containing plasmid induced by the transcription activity (Fig. 2b). Thus, the negative supercoiling generated by the frequent transcription facilitates the quick and robust formation of G4/R-loop in supercoiled DNA but not in the relaxed or linear DNA. Therefore, supercoiled DNA is poised to induce rapid firing of transcription, which in turn triggers G4/R-loop formation that suppresses transcription. Furthermore, we confirmed that transcription-generated supercoiling also stimulates stable G4/R-loop formation by comparing relaxed DNA with linear DNA (Fig. 5h, 5X RNAP with relaxed DNA vs 5X RNAP with linear DNA). These findings suggest that supercoiling stimulates the formation of stable G4/R-loops from both pre-existing and transcription-generated DNA superhelicity.
G4 formation stabilizes the R-loop state.
To understand the mechanism by which G4 stabilizes the R-loop structure, we compared the transcription activity of NT-PQS vs. control non-PQS (Fig. 6a, b) in supercoiled DNA. The overall FRET pattern that emerges from both constructs was compared by generating a heatmap that consists of overlayed single molecule traces that are synchronized either at the time of RNAP + NTP addition (Fig. 6c, d, top) or post-synchronized at the time of long-lived G4/R-loop or R-loop formation (Fig. 6c, d, bottom). While transcription in NT-PQS led to a stable high-FRET state, transcription of a control DNA-induced oscillating mid-FRET state indicated an unstable R-loop formation without G4 (Fig. 6c, d, top). By comparing the time FRET changes occur, we find that NT-PQS DNA accelerates the transition into a stable FRET state. The second set of heatmaps for NT-PQS displays a clear two-step transition from low to mid to high FRET state, consistent with the single molecule trace shown in Fig. 5e. The control, which forms only R-loops, exhibits a fluctuating FRET signal before transitioning to a stable R-loop state or returning to the duplexed DNA (See the representative single-molecule traces in Supplementary Fig. 10). These observations demonstrate a unique feature of NT-PQS that enables G4 formation, stabilizing the R-loop state (Fig. 6d). Consistently, under G4 destabilizing condition (-M+) (Supplementary Fig. 11), NT-PQS exhibits a lower R-loop formation rate and fluctuating FRET transitions, similar to the unstable R-looped state seen in control. These results further emphasize the role of G4 in stabilizing the G4/R-loop structure. Furthermore, treatment of Topo 1 decreased R-loop formation for both constructs, signifying the effect of supercoiling in R-loop stabilization regardless of the G4 formation (Fig. 6e).
G4 and R-loop diminish the transcriptional burst.
Next, we investigated the impact of G4/R-loop structures in transcription initiation. Based on the strand relaxation induced by R-loops, which suppress transcription (Fig. 2), we hypothesized that the stably formed R-loops and G4/R-loops impede RNAP loading onto the promoter. To test this, we adopted our previous FRET configuration suited for monitoring transcription initiation (Fig. 7).39,56,57 [FRET2] construct has Cy3 and Cy5 at 4 bp (-4) and 19 bp (+ 19) downstream from TSS, respectively. The expected FRET value of the immobilized DNA is ~ 0.4, which transitions to ~ 0.7 peak when RNAP opens the transcription bubble and transcribes > 4 nt. (Fig. 7a). Single-molecule traces showed frequent FRET peaks for varying durations (Fig. 7b). Although FRET 2 does not report on the formation of G4/R-loop, we interpret the sudden reduction in initiation events as coincident with the abrupt G4/R-loop formation shown in Fig. 5c-e. The emerging picture is that RNAP loading is accelerated due to the facilitated underwinding of the supercoiled DNA. This will promote R-loop and G4 formation, removing the supercoiling thereby reducing the RNAP loading. To test this expected sequence of events, we quantified the FRET peaks at various stages of transcription, i.e., 200 sec, 10, and 30 minutes after RNAP + NTP addition (Fig. 7c). It shows that transcription initiation is correlated with the formation of G4/R-loop (or R-loop in control) (Fig. 7c). Compared to the control, the NT-PQS exhibits faster G4/R-loop formation, leading to quicker suppression of transcription initiation. The level of negative supercoiling is likely reduced upon the formation of G4/R-loop structures to a level similar to that of a relaxed DNA, which displays a markedly lower initiation frequency, averaging only around 3 events per 200 seconds (Supplementary Fig. 12). In addition, event-driven stochastic simulation58 demonstrated that the promoter supercoiling increases as G4/R-loops (or R-loops in control) accumulate over transcription time. NT-PQS exhibits faster supercoiling increase at promoter than the control (Fig. 7d, and supplementary Figs. 14 and 15).
In conclusion, our investigation highlights the critical role of G4 and R-loop structures in regulating transcription initiation dynamics through modulation of DNA superhelicity at the promoter region. The cooperative formation of G4/R-loop effectively suppresses the heightened transcriptional activity by impeding RNAP loading onto the promoter. Through FRET analysis and event-driven stochastic simulations, we elucidated the underlying mechanisms by which G4/R-loop structures alter DNA supercoiling and promoter accessibility, ultimately influencing transcription initiation rates (Fig. 7e).