Backtracking behavior in viral RNA-dependent RNA polymerase provides the basis for a second initiation site

Abstract

Transcription in RNA viruses is highly dynamic, with a variety of pauses interrupting nucleotide addition by RNA-dependent RNA polymerase (RdRp). For example, rare but lengthy pauses (>20 s) have been linked to backtracking for viral single-subunit RdRps. However, while such backtracking has been well characterized for multi-subunit RNA polymerases (RNAPs) from bacteria and yeast, little is known about the details of viral RdRp backtracking and its biological roles. Using high-throughput magnetic tweezers, we quantify the backtracking by RdRp from the double-stranded (ds) RNA bacteriophage Φ6, a model system for RdRps. We characterize the probability of entering long backtracks as a function of force and propose a model in which the bias toward backtracking is determined by the base paring at the dsRNA fork. We further discover that extensive backtracking provides access to a new 3'-end that allows for the de novo initiation of a second RdRp. This previously unidentified behavior provides a new mechanism for rapid RNA synthesis using coupled RdRps and hints at a possible regulatory pathway for gene expression during viral RNA transcription.

Figure 1.

Experimental configuration of the magnetic tweezers assay and identification of RdRp backtracking. (A) Schematic representation of the experimental assay (not to scale). A predominantly dsRNA construct, attached to the surface by one end and to a micron-sized magnetic bead by the other end, experiences a constant force applied to a magnetic bead (light brown sphere) via a pair of permanent magnets (red and blue cubes) placed above the flow cell (not shown). One of the strands of the construct has a free 3′ end for initiation by a single P2 RdRp. A reference bead (dark brown sphere) is non-specifically adsorbed to the surface to correct for the influence of mechanical drift. During elongation, P2 RdRp synthesizes a complementary copy to the template strand, and in doing so exposes a ssRNA strand between the flow cell surface and the magnetic bead. The resulting change in extension is monitored and converted into a number of incorporated nucleotides. (B) Transcription activity by P2 RdRp versus time monitored using parallelized detection, yielding 52 traces in a single experiment. The experimental conditions include an applied force of 30 pN, an acquisition frequency of 25 Hz (low-pass filtered at 0.5 Hz) and 1 mM ATP/GTP and 0.2 mM CTP/UTP. (C) A sample trace illustrating backtracking behavior by P2 RdRp. A decrease in the length of the transcribed product by ≈15 nts is observed at ≈15 s. Subsequently, P2 RdRp pauses for ≈40 s before resuming elongation. For this trace, the applied force is 35 pN (Materials and Methods). The raw data (blue) are acquired at 25 Hz and filtered at 0.5 Hz (black line). (D) A dwell-time distribution extracted from 52 traces of P2 transcription activity acquired at 20 pN and [NTP]_opt (gray dots). We fit this distribution to a stochastic-pausing model by using MLE (dashed black line)(8). For clarity, we individually plot each contribution to the dwell-time distribution: in green, a Gamma distribution capturing the elongation peak; in blue, the first short exponential pause (Pause 1) and the second short exponential pause (Pause 2); and in red, the power law distribution of pause times originating from backtracking. Above the dwell-time distribution, representative P2 activity events are plotted: from left to right, fast incorporation without pause (light gray), short pauses (light gray) and long backtracked pauses (dark gray). The fit parameters for the Gamma distribution and the exponential distributions extracted from the MLE are available in (8).

Figure 2.

Force dependence of the probability of entering into a backtracked state for P2 RdRp. (A) Probability density distributions for P2 RdRp transcriptional activity acquired at 16 pN (dark blue, 102 traces), 20 pN (light blue, 184 traces), 25 pN (green, 200 traces), 30 pN (yellow, 210 traces) and 35 pN (red, 76 traces). The error bars correspond to one standard deviation estimated from 1000 bootstraps. (B) A zoom-in of (A) for dwell-times longer than 10 s. (C) Probability that a dwell-time exceeds 20 s as a function of the applied force. The error bars are the standard deviation of the distribution extracted from 1000 bootstraps. (D) Proposed model that accounts for the force-dependence of the probability of finding P2 RdRp in a backtracked state. At low force (small distance between single nucleotides on the non-template strand), the tension at the dsRNA fork is small enough to allow the rehybridization of the template strand to the non-template strand, compensating the melting of the dsRNA product, leading to the backtracking of the RdRp. At high force (which results in a larger distance between single nucleotides on the non-template strand), the increase in tension impairs the rebridization of the template to the non-template strand, preventing backtracking to happen.

Figure 3.

Long pauses in the elongation dynamics are rarely followed by processive changes in the direction of transcription. (A) The extension of the transcribed product as a function of time. At 800 s (indicated by arrow), P2 RdRp exhibits a reversal behavior. The experimental conditions include an applied force of 16 pN force and an acquisition frequency of 25 Hz. (B) The probability of observing a reversal event as a function of the applied force (see panel (A)). The error bars represent the 95% confidence interval for a binomial distribution for the ordinate and the standard deviation of the applied force (±5%, see (50)) for the abscissa. (C and D) P2 RdRp transcription activity traces that exhibit very rare behavior in which multiple switches in the apparent directionality of the signal can be observed. In both panels, the extension of the RNA construct (1) increases, then (2) decreases (as in a reversal event) and (3) increases again. Data are acquired at an applied force of 25 pN using buffer conditions described in Materials and Methods. For panels (A), (C) and (D), raw data are shown in blue and data low-pass filtered at 0.5 Hz are shown in black.

Figure 4.

P2 RdRp exhibits different kinetic properties in standard elongation compared to reversal events. (A) The dwell-time distribution for 16 reversal traces obtained at 16 pN (pink circles) and for the forward traces obtained at 35 pN (red circles). Error bars are one standard deviation confidence intervals derived from 1000 bootstraps. (B) The nucleotide addition rates extracted from the maxima of the dwell-time distributions in (A) (identical color codes employed). Error bars are one standard deviation confidence intervals derived from 10000 bootstraps.

Figure 5.

Mechanism accounting for the observation of P2 RdRp reversal events. (A) (Top) The mechanochemical model and (bottom) a schematic trace for a reversal event (Figure 3A). Reversal events are proposed to originate from a second P2 RdRp that initiates on the 3′ RNA product end that extrudes from the NTP channel of a first, backtracked P2 RdRp. The second P2 RdRp is free to initiate template-dependent RNA polymerization on the exposed RNA 3′-terminus. Subsequent elongation will cause the first P2 RdRp to be pushed back along the template strand. (+)RNA is shown in blue and (−)RNA is shown in pink. The background shading represents the different stages of RNA synthesis: gray for the absence of catalytic activity, blue for (+)RNA synthesis and pink for (−)RNA synthesis. (B) Schematic representing RNA production first by the forward transcribing RdRp and then by the second RdRp upon reversal (color code identical to (A)). Strands aligned in parallel fashion are fully hybridized. The net product of a complete reversal event is the original dsRNA construct plus a shorter dsRNA.