Mechanism of sequence-specific pausing of bacterial RNA polymerase

Results and Discussion

Pausing Pattern of E. coli RNAP on T7A1 D111 Template.

The pausing pattern of E. coli RNAP transcribing bacteriophage T7 A1 D111 template determined in our experimental system is fully consistent with published results (29, 30). E. coli RNAP undergoes transient pauses at A26, A31, A32, and C37 positions (Fig. 1, lanes 1–9) at 10 μM NTP. The C37 pause appears predominant. At a higher (200 μM) concentration of NTPs, all pauses decrease, A26 and A31 pauses virtually disappear, and the C37 pause remains predominant (Fig. 1, lanes 10–15). At this site, pausing occurs after incorporation of CTP and before incorporation of ATP. As for the other pauses described (29, 30), the pausing duration depended on the concentration of substrate CTP, with the pause escape rate increasing from 0.032 s⁻¹ at 10 μM CTP to 0.46 s⁻¹ at 200 μM CTP (Fig. S1 A and B). C37 pausing is independent from the RNA secondary structure, because truncation of the nascent RNA in TEC37 by degradation with RNase T1 does not affect pausing in this position (Fig. S1D). Taken together, these observations suggest that C37 represents an example of a sequence-specific ubiquitous pausing site that does not depend on the nascent RNA secondary structure (31). We further characterized properties of the TEC paused at the C37 position to determine the pausing mechanism.

Pausing at C37 Is Not Associated with Backtracking or Translocation Delay.

The extent of backtracking at a given site correlates with the length of time of RNAP stalling by substrate deprivation (16, 18). To address the connection between backtracking and pausing at C37, we tested whether the escape from the pause is slowed down by stalling the TEC at the pause site, as has been reported for backtracked pauses (18, 19). RNAP was set to transcribe through the pause site at 10 μM NTP starting from position C35 (Fig. 2A, read-through) and C37 (Fig. 2A, stalled). In this setup, Pol II from C35 complex was able to proceed through the pause site without artificial stopping. The graphs in Fig. 2A show the percentage of C37 complex as a function of time. Apparently, C37 TEC stalled at the pause site for 10 min and its counterpart that transcribed through the pause site starting from the C35 position exhibited the same dynamics of escape from the pause.

Backtracked pauses and arrests are suppressed by oligonucleotides that hybridize to the transcript 12–20 bp upstream from the 3′ end of the paused/arrested TEC (16, 18). We tested the effect of antisense oligonucleotides hybridized to the 5′ part of 37-nt RNA on the C37 pause by walking the TEC to the A36 position, adding an antisense oligonucleotide, and advancing the TEC to the pause site by addition of CTP. Subsequently, escape from the pause site was induced by addition of all 4 NTPs. Hybridization of short DNA oligonucleotides to the nascent transcript at a distance of 9, 15, and 21 bases upstream from the 3′ end of the RNA resulted in a modest (≈2-fold) decrease of the C37 pause half-life (Fig. 2B). The efficient cleavage of the 37-nt transcript with RNase H confirmed the successful complementary DNA oligonucleotide hybridization to the nascent RNA (Fig. 2B Inset, compare lanes 1 and 2). The absence of sensitivity to stalling or antisense oligonucleotides by the C37 pause is not consistent with a backtracking mechanism for pausing at this site. In addition, pausing at C37 is not decreased by the transcript cleavage factor GreB (Fig. S2B), which reactivates backtracked TECs (16).

Next, we tested whether pausing at the C37 is caused by a delay of RNAP translocation. The pretranslocated and posttranslocated states of the TEC differ in the position of the 3′ end of the transcript relative to the active center of RNAP. The pretranslocated complex contains the 3′ end of the RNA occupying the NTP binding pocket in the active center, preventing loading of the next substrate, which could be a source of pausing. The posttranslocated TEC originates from the pretranslocated TEC by a 1-bp forward shift of the enzyme along the transcript and the template leading to the clearance of the NTP binding site. In the absence of NTPs, the TEC appears to be in equilibrium between these 2 states at each template position, which could be detected by using time-resolved exonuclease III (Exo III) footprinting (28, 32). The rear-end footprint of active A36 appeared as a single band 14 bp upstream from the RNA 3′ end at all times of incubation with Exo III (Fig. 2C, lanes 1–3). The 14-nt distance between the 3′ end of the RNA and the rear-end boundary is characteristic of the posttranslocated TEC (28). Notably, the A36 boundary shifts 1 bp downstream upon formation of the strongly-paused C37 complex (Fig. 2C, lanes 3–6), suggesting that the translocation states of the active A36 and the paused C37 TECs are the same. Importantly, the absence of Exo III degradation products >34 nt, which would reveal C37 backtracking, provide further evidence that RNAP remains engaged with the 3′ end of the RNA in the paused C37 complex. To confirm these conclusions, we performed front-end Exo III footprinting of the A36 and C37 complexes.

At short incubation times with Exo III, the front-end footprint of an active A36 complex represents a double band corresponding to both pretranslocated and posttranslocated boundaries located 17 and 18 bp downstream from the 3′ end of the RNA (Fig. 2D, lane 1). Upon longer incubation with Exo III (Fig. 2D, lanes 2 and 3), the 18-bp boundary disappears, suggesting that the A36 complex easily shifts to the pretranslocated state, providing access of Exo III to an extra base in DNA. In the paused C37 TEC, the footprint shifts 1 bp downstream from its position in the A36 TEC (Fig. 2D, lanes 4–6), demonstrating that the distance between the RNAP boundary and the end of the transcript is the same in the active A36 and paused C37 TECs. Furthermore, the equilibrium between the pretranslocated and posttranslocated states in the active A36 and paused C37 TECs is nearly identical as revealed by the similar rate of disappearance of the posttranslocated boundary upon prolonged incubation with Exo III (Fig. 2D, lanes 2, 3, 5, and 6). Once again, the front-end footprint of the paused C37 TEC does not reveal the presence of backtracked RNAP, which would result in accumulation of products <105 nt. We conclude that pausing at the C37 position is not caused by delayed translocation or backtracking.

Paused TEC Binds Incoming NTP as Efficiently as an Active TEC.

Next, we addressed the mechanism of pausing by testing its sequence determinants. We made a single substitution in the D111 template sequence, changing C37 to G37. This mutation practically eliminated pausing at position 37, leaving the A26, A31, and A32 pauses intact (Fig. 3A, compare lanes 1–15 and 16–30). Therefore, CMP at the RNA 3′ end promoted formation of a paused conformation in the TEC with a 37-nt RNA, whereas GMP at the 3′ end prevented pausing, increasing the rate of the escape from the pause 20-fold (Fig. 3B). Comparison of Exo III footprints of C37 and G37 TECs showed that the equilibrium between the posttranslocated and pretranslocated states has not been affected by the base substitution (Fig. 3C, compare lanes 1–3 and 7–9). Binding of the incoming complementary NTP to a stalled TEC can be detected by a characteristic shift of the pretranslocation/posttranslocation equilibrium to the posttranslocated state as a result of stabilization of the posttranslocated TEC by the NTP in the active center (28). Because the templates carrying paused C37 and active G37 TECs encode AMP in position 38, Exo III footprinting in the presence of nonhydrolyzable ATP analog (AMPcPP) allows comparison of the incoming NTP binding by the paused and active TECs. Fig. 3C shows that paused C37 and active G37 bind an incoming NTP in a similar manner: addition of 1 mM AMPcPP stabilized the posttranslocated conformation of both C37 and G37 TECs (compare lanes 4–6 and 10–12). Quantitative analyses of the enzyme boundary shift induced by AMPcPP (Fig. 3D) confirmed that response of the 2 TECs to the incoming complementary substrate is nearly identical. Therefore, the paused TEC binds the incoming NTP as efficiently as the active TEC does. The delay in the nucleotide addition appears to take place after the incoming NTP enters the active site. Therefore, we addressed the role of the incoming NTP in formation of the pausing intermediate.

Pausing Depends on the Nature of the Incoming NTP.

Three derivatives of the D111 template were generated, in which AMP encoded in position 38 was substituted with CMP, GMP, and UMP. All of these templates encoded CMP at position 37. Pausing patterns on the original D111 template are compared with the 3 mutant variants in Fig. 4A. Evidently, substitution of AMP by GMP only slightly decreases the pausing efficiency, but substitution of AMP to CMP or UMP practically eliminates pausing at C37. Importantly, mutation of AMP to CMP does not change the equilibrium between the pretranslocated and posttranslocated states in the C37 TEC (Fig. 4B), once again arguing against involvement of a translocation block in C37 pausing. Indeed, the dynamics of posttranslocated front-end boundaries in the C37 complex obtained on the original D111 sequence with a terminating 3′ dCMP incorporated at the 3′ end, and the C37 complex obtained in the same way on the D111C38 template are indistinguishable (Fig. 4B, compare lanes 1–3 and 4–6). The dramatic effect of the incoming NTP substitutions on pausing, but not on the translocation equilibrium, is consistent with the finding that paused and active TECs translocate and bind the substrate NTP with similar efficiency. It suggests that sequence-specific pausing analyzed in this work strongly depends not only on the nature of the 3′NMP in the pause site, but also on the nature of the incoming NTP.

Equilibrium Between the Paused and Nonpaused Conformations of the TEC at a Pause Site.

Although the C37 complex on the D111 template displayed a uniform slow rate of escape from a pause (Fig. S1 A and B), the kinetics of escape on the D111U38 template (UTP is the incoming substrate) is noticeably biphasic (Fig. 5A). The escape dynamics are better described by a double exponential function (a fit shown by the bold curve in Fig. 5A) compared with the single exponential function (a fit shown by the thin curve in Fig. 5A). The double-exponential fit suggests that most (70%) of the C37 TEC on the D111U38 template incorporates the substrate with a rate of 0.7 s⁻¹, which at 10 μM substrate is characteristic for an active TEC (Fig. 5A). A small (20%) fraction incorporates UTP with a 0.06 s⁻¹ rate, which is close to the rate established for the paused C37 TEC formed on the standard D111 template. A similar result was observed for CTP incorporation on the D111C38 template coding for CTP as the next NTP after the pause; in this case, the slow fraction was even smaller than that on the D111U38 template (Table S1).

Appearance of fast and slow TECs advancing from the pause site (at a pausing efficiency <100%) indicates that the paused conformation is an intermediate that is not present on the regular elongation pathway (1, 3). The 98% efficiency of pausing at C37 (Fig. S1C) is exceptional, because escape from other pauses on the D111 template appears biphasic. For example, the A32 pausing efficiency is <70% (Fig. S3C), because A32 incorporates the next NTP with biphasic kinetics (Fig. S3 A and B). As in C37 formed on the D111U38 template, the fast rate of NTP incorporation by A32 is characteristic of nonpaused TECs. The slow rate of escape from A32 is 3-fold higher than the rate of escape from the C37 pause. Importantly, pausing at A32 occurs by the same mechanism as pausing at C37, based on the response of the A32 pause to stalling and antisense oligonucleotides (compare Fig. 2 A and B and Fig. S4 A and C). Therefore, the apparently homogeneous C37 pause on the original D111 template should also occur by a branched pathway mechanism.

The presence of the active and paused intermediates in C37 was revealed by changing the catalytic Mg²⁺ to Mn²⁺. Mn²⁺ has a more relaxed coordination requirements compared with Mg²⁺, and substitution of Mg²⁺ to Mn²⁺ changes the conformation of the active site in RNA and DNA polymerases (reviewed in ref. 33). Replacement of the catalytic Mg²⁺ for Mn²⁺ promotes transcription (34), which could be caused by suppression of pausing and/or to an increase of the bond formation rate. Indeed, substitution of Mg²⁺ with Mn²⁺ makes the kinetics of ATP incorporation at the C37 TEC biphasic, with ≈30% of TECs incorporating the ATP 20 times faster than the remaining TEC (Fig. 5B). Notably, both fast (nonpaused) and slow (paused) rates increased significantly compared with the corresponding rates observed in the presence of Mg²⁺ (Fig. 5A and Table S1). The decrease of the pausing efficiency induced by Mn²⁺ suggests that formation of the paused intermediate depends on a specific configuration of the RNAP active site, which may be affected by the nature of the catalytic divalent cations.

The Nature of the Paused Intermediate.

The phosphoryl transfer reaction requires specific geometric configuration of the 2 catalytic magnesium ions associated with the 3′ hydroxy group and the triphosphate moiety of the substrate NTP in the active center of RNAP (33). In RNAP, this configuration depends on the orientation of the 3′ NMP, substrate NTP, and the DNA bases paired to the 3′ NMP and the NTP. Even a subtle difference in stacking of the 3′-terminal RNA–DNA base pair with the base pair of the incoming NTP with the i + 1 base of the template may affect the overall geometry of the active center. Different crystal structures of bacterial and yeast RNAP TECs reveal remarkable variations in the spatial alignment of the incoming substrate and the 2 Mg²⁺ ions with the 3′ terminus of the RNA (35–38). Because only 1 of the described states has a catalytically-competent configuration (36), at least some of the other states may be structurally similar to the paused conformation characterized in our present work. Importantly, the 2 TEC fractions with dramatically different elongation rates observed at the pause sites (Fig. 5 A and B and Fig. S3) suggest that paused intermediate is distinct from the intermediates of the main reaction pathway (31).

The pausing model shown in Fig. 5C is based on the assumption that the incoming NTP might pair with the DNA base, which is not yet fully aligned with the 3′ NMP (36). The resulting intermediate 1 gives rise to at least two distinct conformations (Fig. 5C, intermediates 2 and 3), only one of which is capable of rapid bond formation (intermediate 2). The pathway choice depends on the 3′ NMP and the incoming NTP in the given sequence context. Specifically, 3′ CMP supports the transition of the incoming ATP to the paused configuration (1 → 3) rather than the transition to the catalytically-competent configuration (1 → 2). In contrast, 3′GMP supports the productive alignment of ATP. Unlike ATP, incoming UTP or CTP preferentially form a catalytically competent intermediate when aligned with 3′ CMP.

We cannot exclude the idea that the conformational branching to the paused intermediate occurs before NTP binding (Fig. 5D). Exceptional flexibility of the DNA base encoding the next NTP substrate in the posttranslocated TEC (36) suggests that the DNA base might acquire an active or paused conformations depending on the nature of the 3′ NMP–dNMP base pair. In the case of C37 pause, the dG–rC base pair would support a paused conformation of the next dT base, and an active conformation of the next dA or dG bases. The dC–rG base pair would promote active conformation of the dT base. To be consistent with result shown in Fig. 3, this model requires that the incoming NTP pairs with the paused and active conformations of the DNA base with equal efficiency.

Evidently, a specific combination of 3′NMP and incoming NTP is necessary but not sufficient to induce pausing. E. coli RNAP does not pause at all at C/A or A/U junctions in DNA. The sequences of the RNA–DNA hybrid and downstream DNA affect hairpin-stabilized pauses (3, 27, 39). Similarly, the conformations of the 3′ NMP and incoming NTP in the hairpin-independent pauses might depend on the local sequence context. A detailed comparison of other aspects of the hairpin-independent pausing studied in this work and hairpin-stabilized pausing is provided in SI Text.

Sequence-Specific Pausing Is Not Conserved Between Prokaryotic RNAPs and S. cerevisiae Pol II.

The model for sequence-specific pausing (Fig. 5) is based mainly on the structure of the T. thermophilus RNAP TEC (36). The T. thermophilus RNAP active site shares high sequence homology with E. coli RNAP. These 2 RNAPs are so closely related that even promoter recognition is conserved between them. We analyzed transcription of the D111 template by T. thermophilus RNAP (Fig. 6A) and observed that its pausing pattern is very similar to that of E. coli RNAP (Fig. 1). This finding suggests that the actual structure of the paused intermediate could be resolved by using X-ray crystallography methods.

E. coli RNAP and S. cerevisiae Pol II share the architecture of the RNA-DNA hybrid and the transcription bubble (35, 40–43). Pol II recognizes prokaryotic intrinsic transcription terminator in vitro (44). The catalytic and fidelity mechanisms of the bacterial and yeast enzymes are also similar, relying on the mobility of the trigger loop, a small mobile element, which closes on the cognate NTP in the active site (36, 37, 42, 45). The trigger loop has been recently implicated in hairpin-stabilized pausing of E. coli RNAP (28). At least some sequence determinants of pausing could be conserved between E. coli RNAP and S. cerevisiae Pol II. We compared the transcription patterns by TECs assembled from DNA and RNA oligonucleotides by using E. coli RNAP core enzyme and 12-subunit yeast Pol II on the D111 template.

Contrary to our expectations, the pausing pattern of Pol II is drastically different from the pausing pattern of E. coli RNAP (Fig. 6B). Pol II transcribes the A31, A32 and C37 positions without detectable delay (Fig. 6B, lanes 16–26) with some pausing and irreversible arrest only at the U21 and U27 positions at 10 μM NTP (Fig. 6B, lanes 16–30). U21 and U27 also appear as minor sites of transcriptional pausing and arrest for E. coli RNAP (Figs. 1A and 6B, lanes 1–15). The U27 TEC is severely inactivated by stalling (Fig. S4B). Pausing at this site is eliminated by annealing of antisense oligonucleotides to the upstream segment of the nascent RNA (Fig. S4D). These properties of the U27 complex are consistent with the backtracking mechanism of pausing at this site. Indeed, the U27 TEC formed by E. coli RNAP served as a paradigm of an arrested TEC in a study that established backtracking as a mechanism of transcriptional arrest (16).

The involvement of backtracking in pausing at U27 was confirmed by Exo III footprinting of E. coli RNAP and Pol II TECs (Fig. 6C). The Exo III footprinting revealed a distinct rear-end boundary located 14 bp from the RNA 3′ end of the bacterial enzyme or 15–16 bp from the RNA 3′ end of Pol II (Fig. 6C, lanes 1–3 and 10–12) in the A20 TEC, but failed to detect a corresponding boundary in the U27 TEC (Fig. 6C, lanes 4–9). The RNAP boundary in the U27 TEC appears to be located ≈30 bp upstream from the RNA 3′ end (Fig. 6C, lanes 4–9). Pauses/arrests of Pol II at U21 and U27 positions become less pronounced in the presence of transcript cleavage factor TFIIS (Fig. S2A). In summary, the pauses in common between bacterial and yeast RNAPs occur by a backtracking mechanism, suggesting that backtracking determinants are conserved between the bacterial and yeast enzymes.

The major pauses of Pol II appear to occur by backtracking. However, we cannot rule out that Pol II undergoes backtracking-independent pausing on some sequences that are distinct from the pausing signals recognized by E. coli RNAP. In fact, pausing of Pol II at the C12, C14, and C17 positions of the D111 template are likely to be backtracking-independent, because they are suppressed by a rpb1-E1103G mutation (46) that is proposed to increase the mobility of the trigger loop, but does not promote forward translocation (47). The role of backtracking-independent pausing for Pol II might be less prominent than for E. coli RNAP; the specific regulatory roles of the 2 pausing mechanisms of Pol II remain to be identified.

In conclusion, our findings suggest that backtracking and translocation delay are not the main mechanisms of sequence-specific pausing by bacterial RNAPs. We have established that the paused E. coli RNAP TEC binds NTP as efficiently as the active TEC. We propose that ubiquitous transcriptional pausing involves misalignment of the 3′ NMP and the substrate NTP in the catalytic center of RNAP.