Using mechanical force to probe the mechanism of pausing and arrest during continuous elongation by Escherichia coli RNA polymerase

Effect of Force on Pausing.

Although translocation is not the rate-determining step along the elongation pathway, the strong dependence of the overall rate (which includes pauses) on force strongly suggests that mechanical force can act on the off-pathway reaction leading to a pause (k_P, Scheme S1). Such an effect is consistent with current models for pausing in which a lateral displacement of RNAP relative to the DNA template results in misalignment of the 3′ end of RNA from the catalytic center of the enzyme. Thus, transitions into a paused state should be force-dependent if the rate-limiting step to enter a pause involves this lateral movement of polymerase. To determine the effect of mechanical force on pausing, we consider its effect both on the entry into (k_P) and exit from a pause (k_−P), as can be obtained from the pause density (pauses per kb) and duration. Pauses give rise to minima in the transcription rate; to ensure that these minima correspond to true pauses rather than to artifacts of data sampling and filtering methods, we took only a subset of these minima that met additional criteria, verified by data simulation (see Supporting Text, which is published as supporting information on the PNAS web site, www.pnas.org).

Two different observations suggest an effect of force on entry into the pause state, k_P. First, the mean pause density with assisting force is about two-thirds of that with opposing force (Table 2). As NTP concentration increases, the pause density with assisting force decreases further, showing that closer to saturation, RNAP pauses less frequently, consistent with results from bulk biochemistry. This decrease in pause density (per unit length) with increasing NTPs is not due solely to a decrease in time spent waiting at each nucleotide position: a time-based pause density (pauses per 1,000 s on the main elongation pathway) also decreases with increasing NTP (Table 2), suggesting that NTP concentration directly affects k_P. Second, the fraction of molecules entering any detectable pauses during transcription decreases by more than a factor of two with an aiding force (Table 2). Assuming stochastic entry into a pause, a force-independent pausing mechanism would predict a decrease in the percentage of pause-free runs with an increase in length transcribed. Instead, an assisting force increases both the mean transcription length and the fraction of pause-free runs compared with opposing force, providing additional evidence that an assisting force exerts a considerable effect on entry into a pause. As discussed above, force does not affect the kinetics along the main elongation pathway (k_N, Scheme S1), implying that the reduction in pausing that we observe is not caused by changes in the competing rates of pausing and elongation, but by the direct action of force on the off-pathway transition into the pause state (k_P, Scheme S1).

The effect of force on exit from a pause to return to the main elongation pathway (k_−P) can be studied by examining the pause durations. Again, we have accessed this in two ways. First, we determined the duration of each pause (see Supporting Text), then calculated a mean pause half-life by fitting to first-order decay kinetics. These values for pause half-lives are given in Table 2 and exhibit a slight dependence on force direction. It is difficult, however, to determine precisely the duration of a pause, whereas pause locations are easier to score. Thus, to obtain a more reliable estimate of the pause durations, we used the strongly force-dependent overall transcription rates (v_overall) and the force-independent peak transcription rates between pauses (v_max). The time a given polymerase spends actively transcribing a length L is given by L/v_max; the actual time it takes to traverse this length L in an experiment is L/v_overall. The difference between the active and overall transcription times gives the total time spent off-pathway in a pause state and reflects both the pause density and pause duration; dividing this total time by the number of pauses gives the mean pause duration. These values for pause durations, listed in Table 2, appear independent of force. Thus, it seems that the rate of return to the main elongation pathway, k_−P, is either very weakly or not at all force-dependent.

At first glance, there is an apparent contradiction in the force dependence between entry into and exit from the paused state. Pausing is a reversible process, implying that the forward and reverse reactions traverse the same coordinate. How can traversal of this coordinate in one direction be force-dependent whereas force apparently has little affect on the reverse process? The rate of a force-dependent process scales exponentially with force applied and distance to the transition state (28). With an asymmetrically located transition state, the effect of force on reaction rate can be orders of magnitude stronger in one direction than in the other (see Supporting Text), easily accounting for the difference in force dependence between k_P and k_−P.

Effect of Force on Arrest.

Although RNAP has been shown to backtrack during arrest in stalled complexes formed by NTP deprivation, there is no evidence for backtracking into arrest during continuous elongation with all NTPs present. Previous single-molecule experiments using the same template as in this study showed that in the presence of an opposing load, all RNAP molecules arrested before arriving to the end of the template (10.7 kb from the transcription start site) at both 200 μM and 1 mM NTPs (8). Arrest in these experiments is not due to cumulative photodamage of the enzyme, as the optical trap is turned off as soon as a ternary complex is tethered, before resuming transcription. Given the putative large displacement involved in backtracking, arrest of actively transcribing complexes should display significant force dependence, assuming that the rate-limiting step to entry into arrest involves backtracking of the enzyme over this distance. Specifically, mechanical force opposing transcription should enhance the efficiency of backtracking into an arrested state, whereas assisting force should reduce backtracking and, hence, arrest.

Indeed, mechanically assisting transcription markedly reduces the incidence of arrest compared with transcription subjected to a load (Table 3). Even at 200 μM NTPs, fewer than 60% of the molecules arrest under assisting force, and this value decreases to less than 25% as the NTP concentration increases to 1 mM. The arrest density (total arrests per total length transcribed) also exhibits a significant force dependence (Table 3): the arrest density at 200 μM NTP decreases from 0.87 arrests per kb (opposing force) to 0.30 arrests per kb (assisting force). As mentioned earlier, current kinetic models for transcription elongation suggest that arrests are preceded by pausing intermediates. Given the effect of assisting force in reducing the rate of entry into the pause state, it is possible that the reduction in arrest efficiency with assisting force results simply from reduced pause efficiency and not from a direct effect of mechanical force on entry into arrest (k_A, Scheme S1). To address this possibility, we computed the effect on pause density (pauses per kb, ∝k_P) and arrest density (arrests per kb, ∝k_A) of assisting vs. opposing transcription with force. We found that, at 200 μM NTPs, the pause density with an assisting force is 78% of that with opposing force, whereas the arrest density is decreased much further, to 35% of the opposing force value (Tables 2 and 3). The stronger effect of force on arrest density than on pause density indicates that mechanical force acts directly on the rate of entry into the arrested state (k_A). The strong effect of mechanical force on arrest efficiency can be rationalized if pushing the enzyme in the direction of transcription tilts the reaction coordinate to energetically disfavor backtracking into the arrested state. Furthermore, this strong force dependence implies that entry into arrest involves a displacement of RNAP, consistent with the backtracked arrest state previously described for stalled complexes.

Having shown that applying an assisting force can prevent the arrest of the molecule during continuous elongation, we asked whether it is possible to rescue mechanically an already arrested ternary complex and return it to the main elongation pathway. For purified ternary complexes under standard biochemical reaction conditions, entry into the arrested state is an irreversible process, as arrested complexes cannot resume transcription without the addition of a transcript cleavage factor such as GreB (29, 30). This biochemical irreversibility does not imply that the reaction is mechanically irreversible: force could tilt the reaction coordinate, lowering the energetic barrier sufficiently to allow forward displacement of the enzyme. To test this idea, complexes in which elongation had ceased for at least 10 minutes, i.e., arrested by our definition, were subjected to increasing force in the forward direction. These efforts proved largely unsuccessful, with the enzyme resuming transcription in less than 10% of the cases (3 of 41 attempted rescues).

Is this inability to rescue arrested complexes due solely to the existence of a large energetic barrier to forward displacement? Forces up to 50 pN were applied to the arrested complexes in the above experiments, which, if acting over a distance of 10 bp (the estimated extent of the backtrack; ref. 31), results in ≈40 k_BT of work. (The maximum of ≈50 pN of force available in our instrument is very close to the upper limit of 65 pN that can be applied to double-stranded DNA before it undergoes an overstretching transition to a different structural form; ref. 32.) If the arrested enzyme is simply mechanically displaced, i.e., if the reaction coordinate to arrest involves only a displacement of the enzyme along the DNA, the failure to rescue arrested complexes implies that an exceptionally high energetic barrier must be overcome to resume elongation, which is consistent with the irreversibility of arrest in bulk assays. However, because RNAP can enter into arrest under thermal conditions, the energetic barrier leading to arrest at those template locations must be on the order of at most a few k_BT (Fig. 4). Again, assuming that the reaction coordinate for entry into and exit from arrest is the relative position of RNAP along the DNA template, the free energy difference between the active state and the arrested, backtracked state of the enzyme would be a minimum of ≈35 k_BT. This number is unreasonably large, because it would imply an energetic stabilization upon backtracking by an amount greater than the overall binding energy of RNAP to DNA (≈25 k_BT; ref. 33). This absurd result suggests, instead, that the reaction coordinate for arrest is not simply a displacement of the RNAP along the DNA. Rather, it is likely that a force-independent conformational change of the enzyme or ternary complex occurs after backtracking, and that this conformational change is off the mechanical reaction coordinate, effectively rendering the overall process mechanically irreversible.

What is the nature of this structural reorganization of the ternary complex after backtracking? Zaychikov et al. (34) studied active and arrested ternary complexes and found a significant increase in gel mobility upon arrest, which they interpreted as a change in DNA-bending angle leading to a more extended conformation. Crosslinking studies of arrested and productive elongation complexes have shown a change in interactions between the 3′ end of the RNA and polymerase upon arrest and demonstrated that unwound 3′-proximal RNA feeds into the secondary channel of RNAP (19, 35). It is unlikely that a change in RNA location alone would be sufficient to prevent mechanical repositioning of the polymerase and would account for the large energetic barrier to reactivation of transcriptional elongation. It is more likely that the change in contacts between RNA and the polymerase exerts an allosteric effect on the enzyme, triggering a conformational change away from the active configuration. The rescue of arrested complexes by GreB could be caused as much by the removal of the extruded 3′ RNA as by the removal of its allosteric contact with the polymerase.

Does the backtracking of RNAP automatically induce this structural reorganization of the ternary complex; i.e., are the two components of arrest tightly coupled? In a tight coupling scheme, translocation of the enzyme is accompanied by the simultaneous structural reorganization of the complex. A number of observations suggest that this is not the case. For example, bulk biochemical studies showed that backtracked enzymes are still capable of resuming active transcription at some locations (14). Thus, it is plausible that, upon backtracking, the complex is initially in a displaced intermediate state from which it then undergoes a structural reorganization as described above. This last step should be a mechanically irreversible change that prevents mechanical rescue of the complex to the elongation competent form. From the pause durations observed, we conclude that conversion to a transcriptionally inactive form takes place over the course of minutes at room temperature with the NTP concentrations used in this study (200 μM to 1 mM). To test the model of backtracking followed by mechanically irreversible change, we subjected complexes that had ceased transcription for various amounts of time to varying strengths of assisting force. The results of these rescue attempts are shown in Table 4. Complexes are most likely to be rescued mechanically when they have ceased transcription for less than 5 min. As the delay time is increased before applying greater assisting force on the enzyme, the rescue efficiency decreases, until, finally, for waiting times longer than 12 min, we are unable to rescue any complex mechanically. As controls, the lifetimes of pauses for which we did not increase the force after cessation of transcription were determined. With no increase in initial force, the longest pause observed was 400 s, and the average pause duration was 122 s, whereas with an increase in assisting force, the longest pause rescued was 765 s and the average waiting time before successful rescues was 296 s. Thus, an increase in assisting mechanical force can access and rescue complexes that would otherwise become irreversibly arrested. Most importantly, the ability to rescue an arrested enzyme decreases with waiting time, suggesting that we can mechanically rescue complexes that are paused and possibly backtracked but not those that have undergone an irreversible conformational change into the arrested state (Fig. 5).

Kinetic Cycle of Transcription Elongation.

The results of this study allow us to determine the roles of mechanical and chemical forces on the various steps of the elongation cycle. Fig. 5 shows the kinetic scheme for RNA synthesis, pausing, and arrest given in our introductory remarks, including the role of force and NTP concentration determined in this study. Consistent with earlier studies, we have found the transcription rate to be force-independent but NTP-dependent. A reduction in NTP concentration leads to more frequent off-pathway excursions, manifest in an increased pause density. This result is consistent with the results of bulk biochemical assays (27), in which pause efficiencies are enhanced at lower NTP concentrations, and with the scheme of Fig. 5, in which pausing and elongation compete kinetically. The pause density depends also on the sense (assisting or opposing) of applied force, suggesting that at least a subset of pauses involves displacement of RNAP relative to the DNA template. The stronger effect of force on arrest efficiency lends support to the model of a displaced enzyme in the arrest state and suggests that this displacement is significantly larger than in paused states. The large scale of enzyme backtracking is evidenced by our ability to decrease the incidence of arrest and increase the mean length transcribed by the application of an assisting force and by the fact that even with the smallest of opposing forces (<2 pN), the transcribed length is reduced relative to that with small assisting forces, indicating premature arrest of the polymerase. Interestingly, the process of arrest seems to occur by a bipartite mechanism. First, the enzyme undergoes backtracking into an intermediate state (B, Fig. 5), which can be mechanically restored to elongation. This backtracking is followed by the kinetically slower step of a structural reorganization within the ternary elongation complex. Apparently, the conformational transition to the arrested state cannot be affected by force, as we cannot rescue complexes that have been transcriptionally inactive for longer than 12 min. Therefore, it seems that this arrested state can be restored to the active configuration only by biochemical means, after cleavage of downstream RNA by a factor such as GreB (29, 30).

Elongation Models.

The stability of the ternary complex at the end of the template has important implications for models of elongation. At the last position on blunt-ended DNA, there is no interaction between DNA and regions of the polymerase downstream of the active site (the downstream DNA-binding site in the inchworm model, the protein clamp in the sliding clamp model). Thus, the only interactions maintaining the ternary complex are with and within the RNA:DNA hybrid and with upstream DNA- and RNA-binding sites. The majority (84%) of RNAP molecules that reached the blunt downstream end of the DNA in our assisting-force experiments detached from the template. An analysis of the times spent at the end of the template before dissociation gives a half-life of approximately 40 s, a value comparable to the half-life of ternary complexes at the end of a blunt template determined in bulk biochemical experiments (36). We have found no force-dependent trend to these lifetimes in our study, suggesting that mechanical force exerts no significant effect on the stability of these ternary complexes. Our inability to drastically affect these interactions by the application of an assisting force probably reflects the fact that, in the absence of protein–downstream DNA interactions, lateral forces maintaining the polymerase along DNA are still significant.

We have shown that transcriptional arrest can occur during continuous elongation and at NTP concentrations near saturation, even when a mechanical force is acting to prevent backtracking of RNAP. Nonetheless the incidence of arrest is greatly inhibited by an assisting force and by increasing NTP concentrations. Models of a rigid enzyme suggest that entry into the backtracked, arrested state is driven by hybrid energetics. For example, studies that have modified the hydrogen-bonding within the RNA:DNA hybrid affect its stability and the incidence of backtracking (13). The strong effect of force on arrest suggests that the transition from paused to backtracked complexes involves a large displacement, implying that the enzyme does not populate intermediate states between these two configurations.

The experiments outlined in this paper have demonstrated that the application of force can play a unique role in determining the range of polymerase displacement occurring during transcription. That the enzyme can continue not only to grip the DNA but also to transcribe at an NTP-determined rate, even at such high assisting forces as 50 pN, is an impressive feat. It is still not known what the minimal step size of the enzyme is during transcription elongation, which is a crucial detail for discriminating among various models of RNAP translocation. The ability to observe transcription at high forces has direct implications for obtaining high-resolution information on the dynamics of RNA polymerase. More generally, the ability to access a higher range of forces by assisting a motor enzyme may help in gaining more information about the dynamics of other systems.