Cooperation Between RNA Polymerase Molecules in Transcription Elongation

See allHide authors and affiliations

Science  02 May 2003:
Vol. 300, Issue 5620, pp. 801-805
DOI: 10.1126/science.1083219


Transcription elongation is responsible for rapid synthesis of RNA chains of thousands of nucleotides in vivo. In contrast, a single round of transcription performed in vitro is frequently interrupted by pauses and arrests that drastically reduce the elongation rate and the yield of the full-length transcript. Here we demonstrate that most transcriptional delays disappear if more than one RNA polymerase (RNAP) molecule initiates from the same promoter. Anti-arrest and anti-pause effects of trailing RNAP are due to forward translocation of leading (backtracked) complexes. Such cooperation between RNAP molecules links the rate of elongation to the rate of initiation and explains why elongation is still fast and processive in vivo even without anti-arrest factors.

A high rate of transcription elongation [more than 20 nucleotides (nt) per second] in bacteria and eukaryotic cells (13) occurs despite the presence of numerous potential intrinsic and extrinsic blocks. In vitro, the majority of these blocks (pauses and arrests) are caused by backtracking of RNAP, i.e., reverse sliding of the elongation complex (EC) for one or more nucleotides along DNA and RNA (46). During backtracking, the catalytic site of RNAP loses the 3′ end of the transcript and EC becomes inactivated. Reactivation of EC in vitro either occurs spontaneously because of the reversibility of backtracking (5, 6) or requires specialized factors. One class of these factors induces internal transcript cleavage by RNAP to generate the new 3′ end in the catalytic site. GreA, GreB, and TFIIS are bacterial and eukaryotic members of this group of factors, respectively (7, 8). Another protein from Escherichia coli, Mfd, has been recently shown to use adenosine triphosphate (ATP) hydrolysis to reactivate arrested EC without transcript cleavage (9). Remarkably, neither GreA, GreB, TFIIS, nor Mfd are essential for cell growth under normal conditions (1012), suggesting that more general mechanism(s) must be responsible for efficient transcription elongation in vivo.

In contrast to initiation, when only one RNAP molecule occupies a promoter at a time, the elongation phase can involve multiple RNAP molecules moving one after another along the same DNA molecule. Classical electron micrographs of highly transcribed genes [such as ribosomal (rRNA) genes] look like Christmas trees, displaying strings of queuing ECs tightly packed on DNA (13, 14). Here we tested a hypothesis that a cooperative effort by multiple RNAPs during transcription of the same DNA molecule ensures elongation to be extremely efficient. Using pure E. coli RNAP and various templates with well-characterized arrested sites, we demonstrate that in each case trailing ECs rescue leading backtracked ECs by “pushing” them forward. Taken together, our in vitro and in vivo results suggest a general model of the transcription cycle in which robust initiation ensures rapid and processive elongation.

To determine the overall stimulating effect of trailing ECs on elongation by leading ECs, we first compared the yield of full-length products synthesized by leading EC under single- versus multi-round conditions. The first series of experiments was done in solution with two polymerase chain reaction–generated templates with the same strong T7 A1 promoter but different transcribed sequences (15). In each case, we prepared initial EC that was stalled at position +11 (EC11). Its transcript was radiolabeled with [32P] ATP (0.3 μM) at three positions. EC11 was divided into two batches to receive all four nucleotide triphosphates (NTPs) to 1 mM with or without rifampicin (Rif, 100 μM). The chase reaction continued for 5 min at 37°C. Without Rif, RNAPs that did not participate in EC11 formation could initiate transcription from the same promoter and complete at least one round of transcription. This reaction is viewed as multi-round transcription, because more than one RNAP was able to perform transcription on the same DNA molecule. Because Rif blocks promoter clearance (16), reaction with Rif is considered to be single-round transcription. Fig. 1A shows that the amount of end products accumulated during multi-round transcription on both templates was about twice as much as that during single-round transcription. Virtually all radioactivity in multi-round transcription belonged to the first round performed by leading RNAP, because [32P] ATP was diluted more than a thousandfold with ATP. In the case of Template 2, the run-off product could be detected only under multi-round conditions; in single-round transcription, almost all ECs were halted at the strong arrest site near the end of the template. These observations suggest that efficient transcription by leading RNAP relies on actively transcribed trailing RNAP molecules.

Fig. 1.

Stimulation of transcription elongation by multiple initiation. Step-by-step diagrams of each experiment are shown on top [(see details in the text and (15)]. The size of the transcribed region for each template is indicated in parentheses. (A) Accumulation of the full-sized RNA products during pulse-chase transcription in solution under single-round (lanes 1 and 3) or multi-round (lanes 2 and 4) conditions. aEC and rEC indicate the population of arrested and read-through complexes, respectively. The increase in radioactivity in rEC products (% rEC) is due to the decrease in radioactivity spread over the aEC region. The amount of rEC products obtained during the first round of transcription was taken as 100%. bp, base pairs; Ch, chase (10 min) with 1 mM NTPs; T, terminated transcript; %T, termination efficiency. (B) Accumulation of the full-sized RNA products during pulse-chase transcription in the solid phase. Core, the RNAP #2 core; beads, Ni++-chelating agarose. The partial stimulating effect of Core on run-off production (lane 7) was because of a small contamination of the σ70 subunit.

To address the mechanism of this effect, we switched to a more defined solid-phase transcription system in which each individual step can be controlled (15). Two kinds of RNAP were used, one that carries a 6His tag at the COOH-terminus of the β′ subunit (#1) and another without the His tag (#2). The RNAP #2 holoenzyme was purified from greA/greB E. coli and was free of transcript cleavage factors (10, 15). To confirm that effects observed in the previous experiment are reproducible on any DNA and to better visualize the anti-arrest activity by trailing RNAP #2, we used a third template, which has many prominent pause/arrest sites (15). First, initial EC stalled at position +20 (EC20) was prepared and immobilized on Ni++-chelating agarose beads with RNAP #1 and 10-fold molar excess of Template 3 (Fig. 1B, lane 1). The excess DNA ensured that no more than one RNAP molecule initiated from a single DNA molecule. EC20 was labeled with [32P] cytidine triphosphate at three positions. Beads carrying EC20 were washed several times with high-salt (1 M KCl) and standard (100 mM KCl) transcription buffers to remove unincorporated substrates, free DNA, and unstable complexes (lane 1). Next, EC20 was chased for 1 min in the presence of 100 μM NTPs, then beads were washed to remove NTPs. At this time, all complexes were arrested along the template (lane 2). RNAP #2 was added with or without Rif, followed by the addition of NTPs to 1 mM for another 10 min. Although the high concentration of NTP relieved some pauses, allowing a fraction of EC to form the run-off product (lane 3), the majority of complexes remained arrested. However, the presence of the RNAP #2 holoenzyme, but not its core enzyme, forced most ECs to synthesize the run-off transcript (lanes 4 and 7). Rif abolished this effect (lane 5). These results directly implicate specific initiation and promoter clearance by RNAP #2 as essential for its stimulating effect on RNAP #1 elongation. The presence of RNAP #2 without NTPs did not change the pattern of paused transcripts (lane 6), thus excluding any transcript cleavage activity. Notably, GreB, which was added to about the same molar concentration as RNAP #2, had less anti-arrest effect (lane 9).

These results strongly suggest that trailing RNAP #2 promoted elongation of arrested complexes by translocating them forward. To demonstrate this directly, we carried out stepwise transcription in the solid phase (walking) by RNAP #1 to prepare well-characterized arrested complexes at the defined positions on a fourth template (Fig. 2): +27 (panel 1) and +71 (panel 2). Because of the duplication of the corresponding sequence in Template 4 (Fig. 2A), EC71 and EC27 were identical except for the overall length of their RNA transcripts. Upon brief incubation at 37°C, EC27 backtracked for up to 16 nt (6). Exonuclease III (Exo III) DNA footprinting confirms that EC71 also backtracked for up to 16 nt. RNAP #2 rescued most of EC71, but not EC27 (Fig. 2B). We interpret this result to mean that in the case of EC71, RNAP #2 was free to initiate transcription and eventually translocate arrested EC71 forward. Arrested EC27, however, occluded the promoter after backtracking and prevented promoter clearance by RNAP #2, thus prohibiting its anti-arrest activity (Fig. 3B).

Fig. 2.

The anti-arrest activity of trailing RNAP. (A) The transcribed sequence of Template 4. Unit II is an exact duplication of Unit I. The positions of two arrest sites, +27 and +71, (counting from the +1 start of the transcription) are indicated. (B) Comparison of the anti-arrest effects of RNAP #2 on promoter-proximal (EC27) and promoter-distal (EC71) arrest sites. A step-by-step diagram of the experiment is shown on top [(see details in the text and (15)]. Autoradiographs show RNA products isolated from arrested EC27 and EC71 and intermediate complexes obtained during RNAP walking (supporting online text).

Fig. 3.

The anti-arrest mechanism of trailing RNAP. (A) Monitoring forward translocation of arrested EC71 by Exo III DNA footprinting. A step-by-step diagram of the experiment is shown on top [(see details in the text and (15)]. MEt, β-mercaptoethanol. Panel 1 presents radiolabeled RNA from EC55, EC64, and arrested EC71. Panel 2 shows the protection of a terminally 32P-labeled non-template DNA strand from Exo III digestion. Numbers indicate the absolute length of the digested DNA products as determined by sequencing. (B) Summary scheme of the results of Figs. 2 and 3.

To directly monitor forward translocation of arrested EC, we designed an Exo III footprinting experiment (Fig. 3A). Because reactivation of arrested EC would result in chase and disappearance of the footprint, we developed a method to catalytically inactivate EC without changing its lateral mobility on DNA. We used p-hydroxy-mercuribenzoic acid (HgR), which specifically and reversibly modifies cysteine residues (15). Treatment of EC with HgR caused a complete loss of EC catalytic activity without affecting its stability on DNA. EC can be then fully reactivated by β-mercaptoethanol (panel 1). We treated arrested EC71 with HgR followed by washing to remove HgR. Next, RNAP #2 was initiated on the same DNA, followed by Exo III footprinting (panel 2). Exo III detected the front edge of arrested EC71 at position 146 (the position of active EC55) (panel 2, lanes 2 and 4), indicating that the major population of EC71 backtracked for 16 nt. Challenging EC71 with HgR did not change its footprint (panel 2, lane 5), indicating that HgR itself does not affect the location of EC on DNA. RNAP #2 transcription shifted most of the front edge of HgR-treated EC71 forward (panel 2, lane 6). Under the same conditions, the footprint of non-treated EC71 disappeared (panel 2, lane 7), apparently because of the chase of reactivated complex to the end of the template. These results directly demonstrate forward translocation of arrested EC induced by trailing EC (Fig. 3B).

Our findings suggest a general mechanism that renders transcription elongation fast and processive in vivo (supporting online text). The efficiency of elongation by an RNAP molecule depends on other molecules trailing behind. Indeed, on the basis of results presented here and in previous works (5, 6, 17), backtracking of RNAP seems to occur at virtually any given position. However, the rate and amplitude of EC oscillation strictly depends on local sequence. At certain DNA sites, EC is prone to pause or arrest because of the high probability of spontaneous backtracking and a relatively low probability of forward movement (18). Such probabilities vary markedly even at adjacent nucleotide positions, implying that once an elongating molecule switches to the backtracking state, the trailing molecule would be most likely in the active state at the same moment, shifting the equilibrium of leading backtracked RNAP toward the active state. Thus, the elongation process can be viewed as a group effort by all RNAP molecules within the same transcription unit, which functionally links all steps of the transcription cycle together. Indeed, more robustly initiating, closely spaced RNAP molecules would be during elongation, thus decreasing the probability of intrinsic pauses and arrests at any specific site. Also, with more efficient termination, more RNAP molecules would reinitiate.

To test this general prediction, we measured the rate of elongation in vivo as a function of promoter strength. We used an isopropyl-β-D-thiogalactopyranoside (IPTG)–controllable version of the same T7 A1 promoter that we used in our in vitro experiments, fused to the lacZ reporter gene (Fig. 4A) (15). To compare the rates of transcription initiated from strongly or weakly induced promoters, we induced E. coli cultures with 1 mM or 75 μM IPTG, respectively (15) (fig. S3). The time elapsing between the appearance of a specific hybridization signal from probes complementary to the 5′ and 3′ segments of the lacZ transcript was determined by dot blot hybridization (Fig. 4B) (15). The rate of elongation directly correlated with the rate of initiation. In the fully induced gene, it was ∼62 nt/s, and under the partially repressed conditions, only ∼25 nt/s (supporting online text). These in vivo results confirm our conclusion that the greater the RNAP density, the more efficient and rapid the elongation.

Fig. 4.

Transcription elongation rate as a function of promoter strength in vivo. (A) Schematic representation of the template plasmid (15). Open bars represent the plasmid genes. P/O, the T7 A104/03 promoter/operator site; T, terminator. Two lines indicate probes complementary to the lacZ transcript at the indicated positions relative to the 5′ end. (B) Comparison of the rate of elongation between fully induced (1 mM IPTG) and partially induced (75 μM IPTG) T7 A104/03lacZ (15) (fig. S3). Representative dot blots of the early (1) and late (2) probes were used to generate the induction curves shown. The radioactivity at time 0 (before induction) was taken as the background and subtracted from all values. Arrows mark the point of a steep rise in the hybridization signal, indicating that transcription has reached the corresponding region of lacZ. In each case, the time elapsed between two such points was used to calculate the elongation rate. Numbers represent mean values from three independent experiments.

This cooperation mechanism is expected to work particularly well with highly active genes such as those for stable RNA (rRNA and tRNA) and stress-inducible genes (such as heat shock genes) in prokaryotic and eukaryotic cells. Analysis of rRNA and activated hsp gene transcription demonstrates that RNAP molecules are positioned almost without spacing between (13, 14, 19). The highest rate of transcription elongation observed on those genes (20, 21) is consistent with our model. Furthermore, it was shown that transactivators stimulate transcription elongation by eukaryotic RNAP II (22). This observation provides more independent in vivo evidence favoring a general cooperation mechanism of transcription that links the efficiencies of initiation and elongation.

We believe that, at least in prokaryotes, the cooperation mechanism of transcription is not restricted to RNAP molecules. Because transcription and translation are coupled in bacteria, one can envision that the trailing ribosome acts in a similar way as trailing RNAP, i.e., by “pushing” backtracked ECs forward. In this case, a pulling motion applied by a ribosome to the nascent transcript drives forward translocation. The evidence that the active ribosome can relieve strong pauses that occur within leader sequences of biosynthetic operons (23) supports this idea.

It is important to note that values for various regulatory signals (specific pauses and terminators) that have been measured in vitro in single-round or highly inefficient multi-round assays should be reevaluated under conditions that would better reflect the cooperative nature of elongation in vivo. This is also an important consideration for single-molecule studies in which all conclusions regarding functional parameters of EC have been reached (by definition) based on a single RNAP molecule behavior (24). Pauses and terminators are expected to be less efficient during highly active multi-round transcription. For example, the efficiency of termination at the intrinsic λ tR2 terminator drops by ∼30% upon switching from single- to multi-round transcription (Fig. 1A and fig. S1). Furthermore, it has been shown in vivo that the termination efficiency may be influenced by the efficiency of initiation from the upstream promoters (25). These results are consistent with the model in which backtracking-induced pausing at the T-stretch element of the terminator is critical for termination (26). Hence, suppression of this pause by trailing RNAP would result in compromised termination.

Because the structure and functional organization of cellular RNAP is conserved in evolution (27), the cooperation mechanism of transcription should be general for all types of RNAPs. It explains why known anti-arrest factors GreA, GreB, Mfd, and TFIIS are dispensable in vivo under normal growth conditions (1012). Only rare arrests that occur near the promoter and interfere with initiation (Fig. 3B) and those that trap several RNAP molecules at a time (such as long T-stretches) may require additional anti-arrest factors. Anti-arrest factors may be needed most during stress response, when titration of RNAP by induced-stress genes may compromise the cooperative effect associated with housekeeping-gene transcription. This is particularly important during heat shock—a condition that, itself, promotes backtracking. The temperature-sensitive phenotype of greA/greB cells (10) supports this idea.

In addition to their anti-arrest activity, GreA, GreB, and TFIIS have been shown to help overcome elongational blocks imposed by site-specific DNA binding proteins in vitro and in vivo (28, 29). In those cases, however, a substantial fraction of EC was able to read through the block even without transcript cleavage factors. In a separate work, we demonstrate that trailing RNAPs assist leading EC in passing various roadblocks in vitro and in vivo without any transcript cleavage activity (30). Taken together, these observations indicate that not only intrinsic elongation blocks but also roadblocks caused by DNA binding proteins [such as nucleosomes (31)] can be overcome more efficiently in vivo if RNAP molecules act together.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S3

References and Notes

References and Notes

View Abstract

Navigate This Article