RNA polymerase collisions and their role in transcription

ABSTRACT

RNA polymerases are the central enzymes of gene expression and function frequently in either a head-on or co-directional manner on the busy DNA track. Whether and how these collisions between RNA polymerases contribute to transcriptional regulation is mysterious. Increasing evidence from biochemical and single-molecule studies suggests that RNA polymerase collisions function as an important regulator to fine-tune transcription, rather than creating deleterious “traffic jams”. This review summarizes the recent progress on elucidating the consequences of RNA polymerase collisions during transcription and highlights the significance of cooperation and coordination between RNA polymerases.

Introduction

Genomic DNA is a crowded track shared by numerous motor proteins that carry out fundamental processes such as DNA replication, transcription, and repair. Therefore, collisions among these motor proteins are inevitable, and their conflicts have largely been reported to cause traffic jams that threaten genome stability [1–4]. Here, I focus on discussing the collisions between RNA polymerases (RNAPs) and their regulatory roles in transcription processes. RNAP is a molecular motor enzyme that translocates along DNA and converts DNA to RNA [5,6]. It is highly processive and capable of transcribing thousands of base pairs without disengagement from DNA. Electron micrographs of highly transcribed genes (such as ribosomal genes) displayed strings of densely packed RNAPs [7,8], suggesting multiple RNAP molecules can move one after another along the DNA molecule. This raises the question of what occurs when trailing RNAPs rear-end the leading ones. Furthermore, RNAPs also frequently collide head-to-head since convergently aligned gene units (e.g., antisense noncoding transcripts, genes-within-genes, and convergent gene pairs) are prevalent across the genome in many organisms [9–14]. Thus, questions arise as to the fate of the two RNAPs upon head-on collision and the potential advantages of such ubiquitous and evolutionally conserved gene arrangements.

Elegant biochemical and advanced single-molecule assays have allowed direct and in-depth investigation of RNAP collisions, revealing the fate of the transcription complexes upon head-on and co-directional collisions. Here, I review these advancements and highlight the intricate RNAP coordination in regulating transcription elongation and termination.

Co-directional RNAP collisions

Cooperation of co-directional RNAPs in overcoming pauses and obstacles

By using purified E. coli RNAP and DNA templates containing well-characterized pause sites (e.g., his-operon and trp-leader), Epstein and Nudler [15] showed that the presence of an active trailing RNAP promotes the yield of full-length transcription products, providing the first evidence that co-directional RNAPs act synergistically to overcome sequence-dependent pauses and arrests. Later, they reconstituted transcription ahead of protein roadblocks set using site-specific DNA-binding proteins (e.g., lac repressor) and found that trailing RNAPs enable leading RNAPs to read through the roadblocks [16]. These results support the cooperation model of transcription whereby co-directional RNAPs assist one another to overcome intrinsic pauses and extrinsic obstacles (Figure 1(a)).

Figure 1.

Collisions between co-directional RNAPs.

(a) Co-directional RNAPs assist one another to overcome intrinsic pauses (left) and extrinsic obstacles (right).

(b) Direct “pushing” model or (c) long-distance effect via DNA supercoiling were proposed to be responsible for the cooperation between co-directional RNAPs.

(d) Co-directional RNAPs may act synergistically to deal with numerous barriers by either overcoming them or terminating transcription.

Genomic DNA is packaged by histones in eukaryotes or histone-like proteins in bacteria. To further determine whether cooperation between multiple co-directional RNAPs could facilitate transcription through a nucleosome, Jin et al. [17] utilized a single-molecule unzipping assay to accurately map the positions of two RNAPs and a nucleosome on DNA. They found that a single E. coli RNAP frequently paused when colliding with a nucleosome, whereas the presence of a trailing RNAP assisted the leading one in transcribing through the nucleosome. Kulaeva et al. [18] also reported that multiple eukaryotic Pol II complexes can efficiently overcome the nucleosomal barrier and displace the entire histone octamer but a single Pol II cannot.

Although various transcription factors such as GreA, GreB, and Mfd in bacteria and TFIIS in eukaryotes have been shown to rescue arrested RNAPs and overcome elongational blocks, none of them are essential for cell growth under normal conditions [19–23]. The cooperation model between co-directional RNAPs may explain why elongation is still fast and processive in vivo [24], providing a more general and conserved mechanism for rendering efficient transcription elongation in the cell.

Direct physical contact or torsional stress?

To investigate the mechanism for such cooperation, the conformational state of roadblocked elongation complex (EC) was analyzed either by exonuclease footprinting [25] or single-molecule unzipping assay [17]. They both showed that roadblocked EC adopts a backtracked state albeit with variable backtrack distance. Backtracking – the reverse movement of RNAP along DNA resulting in the disengagement of the 3’ end of RNA transcript – is a fundamental property of RNAP that plays multifaceted roles in gene regulation and genome instability [26]. Backtracked complexes can be rescued by transcript cleavage factors (such as GreA/B in bacteria and TFIIS in eukaryotes) which stimulate the hydrolyzing activity of RNAP removing the extruded 3’ end of RNA [27]. Interestingly, GreB and TFIIS were able but not required to assist the trailing RNAPs in helping the leading RNAPs to overcome such barriers [16,25]. Considering the reversibility of roadblock-induced backtracking, these results led to the hypothesis that the trailing RNAP may help the leading RNAP exit such backtracked state and resume elongation by pushing. Single-molecule studies have shown that both E. coli RNAP and eukaryotic Pol II are able to exert mechanical forces and displace proteins [6,17,28]. Thus, it seems any mechanism that could reduce backtracking and exert forces to push RNAPs forward should facilitate their frequency of overcoming barriers (Figure 1(b)). In addition to the cooperation between co-transcribing RNAPs, there is evidence that transcription-coupled translation in bacteria also enables the trailing ribosome to push RNAPs forward [29–31]. Indeed, direct physical contact between the ribosome and RNA polymerase has been extensively investigated by structural studies [32–34]. Wee et al. [35] recently reconstituted the E. coli transcription-translation coupling system and found that a coupled ribosome could decrease the efficiency of hairpin-stabilized transcriptional pauses, supporting this idea. Notably, they found that the coupled ribosome can accelerate the rate of transcription but also increases rNTP misincorporation, thus reducing transcript fidelity. It will be interesting to investigate whether the cooperation between co-directional RNAPs could also affect fidelity and how to best achieve a delicate balance between efficiency and accuracy.

In contrast to the physical push model, recent experiments [36] and mathematical models [37–40] suggest that DNA torsion or supercoiling can mediate long-distance interaction among co-transcribing RNAPs and is responsible for their cooperation (Figure 1(c)). As the transcribing RNAP moves along the double helical DNA, the downstream DNA becomes more twisted (positive supercoiling), while the upstream DNA becomes less twisted (negative supercoiling), which comprises the so-called twin supercoiled domain model [41]. Since its discovery, DNA supercoiling has been extensively documented to regulate transcription and play substantial roles in many essential cellular processes [41–46]. By monitoring transcription-generated DNA supercoiling and torque buildup in real-time using an angular optical trap (AOT), Ma et al. [47] reported that a single RNAP can be stalled by torsional stress accumulated in negative supercoiled upstream DNA. Chong et al. [48] further showed that DNA positive supercoiling ahead slows down and eventually stops transcription initiation and elongation. However, when multiple RNAPs transcribe the same DNA template, negative and positive DNA supercoils between RNAPs may cancel out [49,50], relieving torsional stress on these RNAPs, thus providing faster translocation than a single RNAP. Supporting this concept, Kim et al. [36] first demonstrated that modulating the native lac promoter strength by varying IPTG concentration and thus the RNAP density along the DNA template does not affect the elongation rate. This result argues the notion of a cumulative effect of RNAPs push on elongation rate, as proposed previously [16]. They further found that turning off a promoter results in apparent slowdown of transcribing RNAPs that are over 2 kb downstream away from the promoter. Such long-distance effect is likely mediated by DNA supercoiling since adding type I topoisomerase (TopA, which removes negative supercoils) can restore normal elongation rate after promoter inactivation. This is conceivable as DNA supercoils indeed can diffuse along the DNA [51]. Their work not only puts forth the idea that the cooperative behaviors of RNAPs emerge from transcription-induced DNA supercoiling, but also provides experimental data that establish the topological effects as a proven fact which is consistent with the extensive body of evidence on topological interactions observed in vivo.

The cooperation between RNA polymerases is an intricate process that impacts the efficiency and regulation of transcription. The two potential mechanisms described above are not mutually exclusive and may coexist in the same cell, depending on the transcriptional unit (e.g., density and efficiency of transcription) and physiological conditions. While direct contacts between RNA polymerase and the ribosome are observed in the context of transcription-translation coupling in vivo [34], the in-situ physical interactions between multiple RNA polymerases on the DNA template during transcription have not been directed documented in the current literatures. Advances in techniques that enable simultaneous monitoring the precise locations of multiple RNA polymerases and the dynamic formation of DNA supercoiling under physiological conditions are awaited to further dissect the cooperation mechanisms between RNA polymerases.

Co-directional RNAPs in regulating transcriptional termination

While cooperation between co-directional RNAPs facilitates transcription elongation, it may cause inefficient termination events. Transcriptional termination sites usually harbor hairpin-forming sequences or T-stretch elements that stall elongating RNAPs preparing for termination [52]. The rear-end collisions between co-directional RNAPs could stimulate readthrough over these pause-site, which would result in compromised termination. Indeed, the termination efficiency at an intrinsic tR2 terminator decreased by almost 30% under multi-round transcription conditions when compared to single-round transcription [15]. In vivo studies also showed that the termination efficiency was influenced by the upstream promoter strength [53].

In contrast to inhibiting transcriptional termination, Wang et al. [54] recently found that co-directional RNAP collisions facilitate the dissociation of head-on collided transcription complexes, thus promoting the termination efficiency at the overlapping terminators shared by convergent gene pairs (see detailed explanation in the following section). This is reminiscent to an earlier study reporting that the rear-end collision between bacteriophage T7 RNAPs result in displacement of a downstream stalled elongation complex [55], indicating co-directional RNAPs may promote pre-mature termination when encountering strong barriers.

During transcription, RNAP is distinctly stalled by various pause sequences, DNA lesions, numerous specific or nonspecific DNA binding proteins, and misincorporated substrates [56,57]. It is still mysterious how co-directional RNAPs sense different obstacles and deal with them by either passing through them or terminating transcription (Figure 1(d)).

Head-on RNAP collisions

Head-on RNAP collisions in transcriptional interference

The outcome of head-on RNAP collisions has been studied by atomic force microscopy (AFM), in which Crampton et al. [58] captured the locations of two E. coli RNAPs on a linear DNA template with two convergently aligned promoters. Through time-course analysis of the snapshots of RNAP positions along the DNA upon transcription re-initiation, they found that a significant proportion of the collided elongation complexes remain bound to the DNA and are stalled in close proximity to each other. Unlike the highly stable elongation complex, the promoter-occupied polymerase may be dislodged by colliding with an opposing transcribing polymerase [59]. Similarly, Hobson et al. [60] reconstituted two yeast Pol II in a head-on configuration and reported that Pol II molecules cannot transcribe past one another, suggesting the opposing elongation complex represents strong barriers to each other. Head-on collision of two RNAPs impede transcription progression in both directions, and therefore convergent gene arrangements might be used to regulate gene expression (Figure 2(a)). For example, convergent transcription interference was shown to regulate the lytic-lysogenic switch controlled by face-to-face arranged lytic promoter (pR) and lysogenic promoter (pL) in the non-lambdoid coliphage 186 [61]. Increasing evidence showed that convergent gene pairs negatively affect each other’s expression, such as antisense RNAs (asRNAs) whose amount were inversely correlated to the expression level of related coding genes [59,62,63]. These results suggest an evolutionarily conserved strategy for modulating transcription by gene arrangement and promote the notion that polymerase collisions may be an underlying mechanism that contributes to the impact of antisense transcription.

Figure 2.

Collisions between head-on RNAPs.

(a) Head-on collision of two RNAPs impedes transcription progression in both directions by either direct contact or accumulated torsional stress. Heterogeneously collided RNAPs could be removed by ubiquitin-proteasome system or other mechanisms such as TC-NER to avoid accumulation of persistently arrested RNAP.

(b) Placement of bidirectional terminator element ensure head-on RNAP collisions occur uniformly at the terminator site, thus maintaining transcript boundaries for convergent genes. Additional co-directional RNAP collisions facilitate the dissociation of head-on collided complex and enable efficient bidirectional termination. Such collision-driven termination was probably achieved either through direct physical interaction or via torsional stress accumulated in the DNA.

Regarding the interference mechanism of head-on RNAPs, both physical contact and long-range torsional stress have been suggested. By analyzing the length of RNA products, Hobson et al. [60] mapped the precise position of the stalled RNAP upon head-on collision and found that the transcribing polymerases stopped when they encountered each other (i.e., physical contact). This close approach relied on a halted RNAP with short RNA transcript and thus is likely much more able to rotate than actual transcription complexes in cells. However, Crampton et al. [58] observed the RNAPs approach each other but did not always reach close contact, suggesting the torsional stress generated in front of each RNAP may account for their stalling. Further investigation is needed to distinguish these mechanisms of RNAP head-on collision.

Head-on RNAP collisions in maintaining transcript boundaries for convergent genes

The studies described above were performed on DNA templates with two promoters separated by a coding region lacking transcriptional terminators. However, using simultaneous 5’ and 3’ end RNA sequencing (SEnd-seq), Ju et al. [64] discovered that there exists a new and prevalent class of bidirectional transcription terminators between convergent gene pairs in E. coli. Convergent transcription (i.e., active transcription events from both directions) is required for efficient bidirectional termination both in vitro and in vivo, suggesting that the head-on conflicts between opposing transcription machineries may offer an important mechanism for terminating transcription at these sites. To test this hypothesis, Wang et al. [54] established a single-molecule platform that integrates both fluorescence detection and force manipulation to allow direct visualization of multiple RNAP trafficking and collisions as well as their compositional changes. By simultaneously tracking both RNAPs and their respective RNA products, they found that the converging RNAPs collide at the overlapping termination site, which efficiently prevents transcriptional readthrough into the opposite gene, but is not sufficient for the release of the RNAPs and nascent RNA. Unexpectedly, an additional trailing RNAP running into the collided complex can drive the last step of termination (i.e., dissociate the head-on collided complex from the DNA template). As such, head-on and co-directional RNAP collisions together orchestrate bidirectional transcription termination (Figure 2(b)). These results highlight an underappreciated role of RNAP conflicts in shaping transcript boundaries.

Although the DNA sequence of the bidirectional terminator element does not cause strong termination per se, it encodes the RNA hairpin structure that could frequently trap RNAPs from either direction [54]. Deletion of the bidirectional terminator element results in heterogeneous collision sites, suggesting it serves as an important pausing signal for synchronization between the converging RNAPs and ensures their collisions occur uniformly at the terminator site [54] (Figure 2(b)). Therefore, the placement of bidirectional terminator elements between convergent genes is critical to shaping the 3’ boundaries of transcripts. Given the ubiquitous existence of convergent gene pairs [9–14], future studies are needed to assess the prevalence of similar elements and RNAP collision-driven termination in other species.

Resolve head-on collide-arrested RNAPs

Since persistently stalled transcription complex may constitute a serious threat to genome stability [3,65], cells must develop strategies to resolve these potential disastrous consequences. To deal with stalled Pol II caused by different obstacles, eukaryotic cells have invoked distinct mechanisms including TFIIS-dependent RNA cleavage and backtrack rescue, premature transcription termination, and degradation of polyubiquitylated Pol II [57]. Bacterial cells also harbor effective proteins (such as Mfd and Rho) functioning in traffic clearance and conflict resolution during transcription process [66–69]. In yeast, the head-on collided Pol II has been shown to trigger Pol II polyubiquitylation and degradation [60] (Figure 2(a)). Whether other mechanisms such as transcription-coupled nucleotide excision repair (TC-NER) [57,70–73] also contribute to removing collide-arrested RNAPs remains elusive. It is also interesting to investigate whether co-directional RNAPs collision could also displace head-on collided RNAPs in the absence of bidirectional terminator elements, thus serving as an alternative strategy to resolve collide-arrested RNAPs.

Discussion and future outlook

In this review, I discussed different types of RNAP collisions and their consequences, highlighting the importance of RNAPs acting together. These advancements have reshaped our understanding about the collisions among RNAPs – typically thought to be deleterious to genome stability and evolutionarily disadvantageous – can in some contexts be harnessed to control the output of gene expression. Collisions between RNAP and other DNA-based motors – most notably the replisome – have been shown to accelerate the mutagenesis and gene evolution [74,75], which could be another physiological function of such collisions. It is worth noting that two recent studies [76,77] have suggested that RNA Pol II may jumpstart transcription in the post-replicative region. Specifically, the evicted Pol II, resulting from replication fork progression, may temporarily associate with the replisome and swiftly resume transcriptional activity upon passage of the replication fork. This would lead to altered RNA synthesis and alternative splicing events that may promote cell fate switches during development or cellular differentiation, adding to the functional roles of their collisions. These studies also suggest that RNA polymerase could potentially be bypassed by other motor enzymes without premature termination, reminiscent of the bypass ability exhibited by phage RNAPs during head-on collisions [78]. Therefore, the bypass of RNA polymerase when encountering obstacles may offer an additional mechanism for cellular regulation.

Recent single-molecule studies also showed that RNAP can be recycled and undergo diffusion after termination, or even flip its orientation and reinitiate transcription [79–81]. Reinitiation may provide a mechanism to orchestrate the transcriptional activities of groups of nearby operons [82]. Therefore, transcription should not be considered as individual initiation-elongation-termination cycles in isolation, but rather a network of RNAPs that need to cooperate and coordinate with each other. How their behaviors are mutually regulated and well-coordinated when transcribing on different cis elements (e.g., varied strength of promoters and terminators, convergent and divergent [83–86] gene arrangements, DNA lesions) and encountering diverse trans factors (e.g., DNA-binding proteins and RNAP-interacting factors) must be further explored. The single-molecule techniques that enable real-time observation of multiple-motors trafficking on DNA [87,88] can be used to investigate a wide range of genomic conflicts and offer dynamic information for such coordination, revealing their potential physiological functions. In combination with structural approaches and single-molecule platforms for monitoring DNA supercoiling [89–91], future studies also need to clarify the molecular state of the collided RNAPs under different circumstances and the exact role of DNA torsional stress during these processes.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).