Nascent RNA structure modulates the transcriptional dynamics of RNA polymerases

Abstract

RNA polymerase pausing represents an important mechanism of transcriptional regulation. In this study, we use a single-molecule transcription assay to investigate the effect of template base-pair composition on pausing by RNA polymerase II and the evolutionarily distinct mitochondrial polymerase Rpo41. For both enzymes, pauses are shorter and less frequent on GC-rich templates. Significantly, incubation with RNase abolishes the template dependence of pausing. A kinetic model, wherein the secondary structure of the nascent RNA poses an energetic barrier to pausing by impeding backtracking along the template, quantitatively predicts the pause densities and durations observed. The energy barriers extracted from the data correlate well with RNA folding energies obtained from cotranscriptional folding simulations. These results reveal that RNA secondary structures provide a cis-acting mechanism by which sequence modulates transcriptional elongation.

Fig. 1.

Single-molecule transcription. (A) Experimental setup. Two optical traps (gray) were used to trap polystyrene beads (brown) functionalized with antidigoxygenin (AD) and streptavidin (SA). A ∼4 kb-long DNA tether was formed between a biotinylated polymerase (blue) and the downstream DNA that contained a 5′ digoxygenin. (B) Representative traces of nuclear (Pol II) and mitochondrial (Rpo41) RNA polymerases of Saccharomyces cerevisiae on GC-rich and AT-rich templates. Data averaged at 50 Hz is shown in gray, Savitsky–Golay filtered data (1 Hz) in black, and pauses in red. (C) Pause-free velocities of Rpo41 and Pol II on different templates. (D) Mean pause duration for the different templates used in this study, as a function of transcript GC composition. The transcript GC composition is run-length weighted, and therefore differs between enzymes due to slightly different run lengths. (E) Mean pause density for the different templates and enzymes used in this study. Unless otherwise noted, error bars are standard errors of the mean. The total number of traces for each condition are given in Table 1.

Fig. 2.

Pause dependence on RNA hairpin structure. (A) Cumulative distributions of pause durations of Pol II on AT-rich (red) and GC-rich (blue) with (solid) and without (dashed) RNase A present in the reaction the buffer. Transcription on the random template is omitted for clarity (Fig. S2). (B) The mean pause duration for Pol II on all templates in the presence of RNase A (dashed line) and in its absence (solid line). (C) The mean pause densities for Pol II on all templates with and without RNase A in the reaction buffer. The total number of traces for each condition are given in Table 1.

Fig. 3.

Template and polymerase dependence of kinetic properties. (A) A schematic of the proposed model, in which the backtracked polymerase (blue) occupies a periodic potential landscape (black) in which the sequence dependence increases the energy of the barrier to backward movement (ΔG_RNA). The forward and backward diffusion rates are indicated as arrows above the transition states upstream and downstream of the polymerase, respectively. (B) The cumulative distribution of pause durations on AT-rich (red) and GC-rich (blue) templates for Pol II (top) and Rpo41 (bottom). Solid lines are experimental data and dashed lines are fits of the model discussed in the text. Transcription on the random template is omitted for clarity (Fig. S2). (C) The correlation between the fitted energy barriers to backtracking (ΔG_RNA) and those calculated via a cotranscriptional folding simulation (ΔG_sim). The horizontal error bars indicate the standard errors of the simulated energies using different seeds for the simulation, and they are smaller than the data makers for the AT-rich and random templates. A fit of the data to a line of unity slope is shown (dashed).