|Home | About | Journals | Submit | Contact Us | Français|
RNA polymerase is a ratchet machine that oscillates between productive and backtracked states at numerous DNA positions. The amount of backtracking (reversible sliding of the enzyme along DNA and RNA) varies from one to many nucleotides. Since its first description 15 years ago, backtracking has been implicated in a plethora of critical processes in bacteria and eukaryotic cells. Here we review the most fundamental roles of this phenomenon in controlling transcription elongation, pausing, termination, fidelity, and genome instability. We also discuss recent progress in understanding the structural and mechanistic properties of the backtracking process.
The first systematic biochemical analysis of transcription elongation complexes (ECs) revealed surprisingly irregular DNA footprints, suggesting that RNA polymerase (RNAP) shrank and expanded during elongation (Krummel and Chamberlin, 1992). This led to a provocative “inchworming” model in which the two-stroke mode of RNA synthesis coupled to a leap-like movement of RNAP constituted the mechanism of transcript elongation (Chamberlin, 1994). Subsequent probing of many ECs stalled in succession over long stretches of DNA revealed that “inchworming” is not obligatory for elongation (Nudler et al., 1994); the irregularities of footprints occurred only at certain DNA sites, whereas the majority of DNA positions displayed relatively monotonic movement of RNAP where each nucleotide addition was accompanied by a one base-pair translocation (Nudler et al., 1994; 1995).
In 1997, two sets of biochemical data showed that occasional “inchworming” was in fact reversible sliding of ECs along DNA and RNA (Komissarova and Kashlev, 1997a; Nudler et al., 1997). One approach used short antisense oligonucleotides complementary to the segment of nascent RNA just emerging from RNAP (Komissarova and Kashlev, 1997a). These oligos diminished the footprint irregularities at the “inchworming” sites, demonstrating that restricting RNA from threading back into the enzyme prohibits oscillation of RNAP. Similar results were obtained when RNA:DNA hybrid stabilizing NTP analogs were used (Nudler et al., 1997). Moreover, analogs that destabilized RNA:DNA hybrid induced “inchworming” at sites where it previously did not exist. The conclusion from these experiments was that the stability of RNA:DNA hybrid is the key determinant of the lateral mobility of the EC. This work also determined the actual length of the RNA:DNA hybrid in the EC to be 8±1 base pairs and introduced the term “backtracking” to define the phenomenon of spontaneous sequence-dependent back and forth sliding of the EC and to distinguish it from “inchworming” (Nudler et al., 1997).
Even though the inchworming model was a misinterpretation of available data, it sparked intensive studies leading to the discovery of backtracking. Moreover, the idea of fixed and moving parts of RNAP first postulated by the inchworming model turned out to be visionary in the context of the ratchet mechanism of RNAP (Bar-Nahum et al., 2005; Brueckner and Cramer et al., 2008; Tagami et al., 2010).
During backtracking the catalytic site becomes disengaged from the 3′ end of RNA, rendering the EC inactive, but stable (Nudler et al., 1997; Komissarova and Kashlev, 1997b) (Figure 1). This disengagement constitutes the mechanistic basis for many regulatory pauses and arrests. Although some strong pauses do not involve extensive backtracking (Toulokhontov et al., 2007; Kireeva and Kashlev, 2009), fraying of the 3′ RNA terminus, that interferes with nucleotide binding and incorporation, was proposed to constitute the initial or elemental pause, from which both backtracked and non-backtracked pauses originate (Toulokhontov et al., 2007; Sydow et al., 2009). Because backtracking involves synchronized rewinding of the hybrid upstream, unwinding/rewinding of the DNA duplexes ahead and behind of a transcription bubble, and threading of the single stranded RNA through RNAP (Figure 1), the overall sequence context determines the probability of pausing at each nucleotide position (Tadigotla et al., 2006).
Promoter-proximal pauses or arrests constitute one large class of backtracking events. They usually occur within the first ~50 nt of transcribed sequences. Several features make promoter proximal regions particularly prone to backtracking: (i) persistent contacts between the enzyme and initiation factors and/or promoter DNA tend to “pull” ECs backward; (ii) nascent RNA is not long enough to form structures that would prohibit reverse sliding; (iii) there is little room for trailing ECs (and in bacteria, ribosomes) to “push” backtracked ECs forward (Epshtein and Nudler, 2003; Proshkin et al., 2010). In addition, regions of specific promoter-proximal sequences also contribute to backtracking (Perdue and Roberts, 2010).
Many E. coli and coliphage operones display early arrests in vitro and in vivo that rely on the action of the sigma initiation factor (σ70) during elongation (Ring et al., 1996; Brodolin et al., 2004; Nickels et al., 2004; Hatoum and Roberts, 2008; Stepanova et al., 2009). Although the functional role for most of these pauses is unknown, σ70-mediated backtracking is required for phage λ Q antitermination. Q protein is recruited to such arrested ECs and together with the elongation factor NusA forms an “antitermination shield” that insulates the moving EC from intrinsic and Rho dependent terminators (Roberts et al., 1998; Shankar et al., 2007).
Promoter-proximal pauses by RNAP II are widespread in mammalian and Drosophila DNAs, particularly in highly active and regulated genes (Zeitlinger et al., 2007; Core et al., 2008; Rahl et al., 2010; Nechaev et al., 2010). These long-lasting pauses play a critical role in the transcriptional regulation of many genes (Bentley and Groudine, 1986; Sawado et al., 2003; Saunders et al., 2006) and also in the processing of RNA transcripts (Glover-Cutter et al., 2008). One of the most well-studied examples of promoter-proximal pausing occurs at the Drosophila heat shock genes (Rougvie and Lis, 1988; Lis, 1998; Wu et al., 2003). ECs located at the promoter of these genes pause as a result of backtracking (Adelman et al., 2005), poising them for rapid reactivation of transcription in response to stress. They also compete with nucleosomes at these highly regulated promoters, thereby inhibiting the formation of repressive chromatin structure and facilitating the rapid resumption of transcription (Gilchrist et al., 2010).
In eukaryotic and prokaryotic cells backtracked (arrested) complexes can be rescued by the transcript cleavage factors TFIIS (Izban and Luse, 1992) and GreA/GreB (Borukhov et al., 1993), respectively. These factors stimulate intrinsic hydrolyzing activity of RNAP, which removes the 3′ extruded portion of the transcript to generate a new RNA 3′ end in the catalytic site, thereby reactivating the EC. The transcript cleavage factors have been shown to relieve promoter proximal pausing (Marr and Roberts, 2000; Adelman et al, 2005; Stepanova et al., 2009). They are particularly important in proximity to the promoter, where other anti-backtracking mechanisms (see below) are limited.
Deep sequencing of the 3′ ends of nascent transcripts associated with yeast RNAP II revealed that backtracking-mediated pausing occurs not only near promoters, but ubiquitously throughout the transcribed sequence of any given gene in vivo (Churchman and Weissman, 2011). Among 2×105 pause sites detected in yeast genome, 75% were associated with backtracking. Indeed, functionally significant backtracking signals occur anywhere within a transcription unit. One classical example is the oligo-T stretches that play a key role in transcription termination (discussed below). Another notable example is the operon polarity suppressor (Ops) element, which is required for activation of some fertility and virulence operons in bacteria (Bailey et al., 1997). Ops is a backtracking type pause signal (Artsimovitch and Landick, 2000) that is required to recruit the antitermination factor Rfh to the EC (Bailey et al., 2000). As discussed below, transient pauses that occur at numerous sites due to spontaneous backtracking control the overall elongation rate and level of gene expression.
In contrast to the initiation step when only one RNAP molecule occupies a promoter at a time, elongation often involves multiple RNAPs moving one after another along the same DNA duplex. Each molecule behaves as a powerful ratchet machine (Bar-Nahum et al., 2005; Tagami et al., 2010) exerting ~20 pN of force (Wang et al., 1998). The probability of backtracking varies dramatically at different positions, even at adjacent nucleotides, implying that when the leading EC backtracks the trailing EC would be most likely in the active mode, “pushing” the leading EC forward. Such cooperation between ECs has been demonstrated in vitro and in vivo for E.coli and yeast RNAPs (Epshtein and Nudler, 2003; Epshtein et al., 2003; Jin et al., 2010, Saeki and Svejstrup, 2009). Therefore, the elongation phase of the transcription cycle should be considered as the effort of the entire group of RNAP molecules within the same transcription unit, which effectively links the initiation and elongation steps; the more robust the initiation, the more closely spaced the elongating RNAP molecules, and thus, the lower the probability of backtracking at any individual position. Indeed, as observed, the stronger the promoter is, the faster the elongation occurs along the transcription unit in vivo (Epshtein and Nudler, 2003). This cooperative mechanism explains, at least in part, why the most active genes, such as stable RNA genes (e.g. rRNA genes) or stress-inducible genes (e.g. heat shock genes) have the highest elongation rates in bacteria and eukaryotes.
This implied cooperation between different RNAP molecules likely extends beyond matching RNA output to promoter strength. The elongation rate modulates alternative splicing by affecting the timing at which splice sites are exposed to the splicing machinery (de la Mata et al., 2003; Kornblitt, 2007). Moreover, paused ECs near the 3′ splice site facilitate co-transcriptional splicing in yeast (Carrillo Oesterreich et al., 2010; Alexander et al., 2010). Thus, RNA processing can be influenced by the robustness of initiation according to the cooperative anti-backtracking mechanism. This kinetic mechanism could work independently or in cooperation with more specific (factor-dependent) mechanisms (Nagaike et al., 2011) in coupling transcription activation to pre-mRNA processing.
In bacteria transcription and translation are coupled. While moving along coding sequences the EC is closely followed by translating ribosomes. As a result of this coupling, the trailing ribosome is able to “push” backtracked RNAP forward, thereby accelerating its speed (Proshkin et al., 2010, Burmann et al., 2010). This “cooperation” between ribosome and RNAP explains how the rate of transcription elongation perfectly matches the rate of translation under various growth conditions (Proshkin et al., 2010). It also explains why it depends on codon usage, i.e. the frequency of rare codons, which modulate the speed of a ribosome (Proshkin et al., 2010). The implications of these findings are far reaching. Not only does this cooperation conserve energy by limiting any excessive transcripts that cannot be translated in a timely manner, it also prevents premature Rho termination by ensuring continuous coupling between transcription and translation. Thus, bacteria rely on trafficking and cooperation to finely control the expression of each individual gene in response to nutrient availability and growth phase.
RNAP must traverse numerous potential roadblocks in vivo, such as nucleosomes (or nucleoid-associated proteins in bacteria) and a plethora of other site-specific and non-specific DNA binding proteins. Despite the constant presence of such roadblocks within intragenic regions, RNAP progresses relatively unimpeded in vivo, although in vitro, protein roadblocks readily inhibit transcription (Izban and Luse, 1991; Reines and Mote 1993; Espshtein et al., 2003; Walter et al., 2003; Lewis et al., 2008). Backtracking resolves this apparent paradox. Even though RNAP exerts sufficient force (~20 pN) (Wang et al., 1998) to displace many DNA-bound proteins in its path, its propensity to backtrack complicates this displacement for any individual molecule. Indeed, upon collisions with different DNA-bound proteins, such as the lac repressor, hydrolytically defective endonuclease EcoRI (E111Q), or a nucleosome, RNAP backtracks in vitro and in vivo (Epshtein et al., 2003; Walter et al., 2003; Churchman and Weissman, 2011). In vitro, it can remain backtracked in front of the roadblock indefinitely (Epshtein et al., 2003). Thus to overcome various roadblocks, RNAP must first be reactivated. Indeed, transcript cleavage factors have been shown to facilitate readthrough of the roadblocks(Reines and Mote 1993; Epshtein et al., 2003; Walter et al., 2003). Moreover, cooperation between RNAP molecules and between RNAP and ribosomes (in bacteria) is a general and efficient mechanism of traversing the roadblocks, including nucleosomes (Epshtein et al., 2003; Proshkin et al., 2010; Jin et al., 2010).
Backtracking provides the basis for the mechanism of transcriptional proofreading. As mentioned above, backtracking strongly depends on the stability of the RNA:DNA hybrid in the transcription bubble and also on the nature of the 3′-terminal residue (Nudler et al., 1997; Sosunov et al., 2003). The weaker the hybrid, the higher is the probability of backtracking (Nudler et al., 1997). Therefore, any mismatch would induce immediate backtracking, which, in turn, would result in cleavage and removal of the 3′ RNA portion that contained a misincorporated nucleotide. Indeed, the GreA transcript cleavage factor and its eukaryotic analogue, SII, have been shown to substantially enhance transcriptional fidelity in vitro (Erie et al., 1993; Jeon and Agarwal, 1996; Thomas et al., 1998) and in vivo (Koyama et al., 2003). Also, the rpoB9 subunit of RNAP II, which facilitates SII-dependent transcript cleavage, contributes to fidelity in vivo (Nesser et al. 2006; Koyama et al. 2007). The mismatch during NTP insertion can also stabilize a paused state of RNAP with a frayed RNA 3′ nucleotide that inactivates RNAP and promotes backtracking and proofreading (Toulokhontov et al., 2007; Sydow et al., 2009).
Other evidence linking backtracking to fidelity comes from the biochemical analysis of RNAP mutants that alter its propensity to backtrack. Backtracking-prone (slow) RNAP usually exhibits a lower rate of nucleotide misincorporation, whereas backtracking-resistant (fast) RNAPs appear to be more error-prone (Bar-Nahum et al., 2005; Kireeva et al., 2008). Binding of the correct NTP in the i+1 site of the catalytic center stabilizes RNAP in the post-translocated state and suppresses backtracking (Bar-Nahum et al., 2005), most likely via substrate-induced folding of the trigger loop domain. By closing around the active center, the trigger loop transiently captures the correct substrate (Vssylyev et al., 2007; Kaplan et al., 2008; Kireeva et al., 2008). At the same time, it partially occludes the secondary channel (NTP delivery pore) through which RNA is extruded during backtracking (Korzheva et al., 2000). Thus, RNAP backtracking, which depends on the trigger loop conformational state (discussed below), may assist in substrate selection; an incorrect NTP facilitates backtracking and hence its own expulsion through the secondary channel, while the correct NTP stabilizes the enzyme in the catalytically competent (backtracking-resistant) mode, thereby facilitating its own incorporation.
Intrinsic termination signals in bacteria consist of a GC-rich inverted repeat element followed immediately by a stretch of T bases (“T-stretch”). The resulting transcript forms a stable hairpin structure followed by several Us at the 3′ terminus. As discussed above, T-stretches are typical backtracking signals because they create weak U:A base pairs in the transcription bubble. Irrespective of the termination hairpin sequence, the T-stretch induces a brief pause precisely at the termination position (Gusarov and Nudler, 1999; 2001). This backtracking-type pausing depends on the 3′ proximal portion of the T-stretch and can also be affected by bases immediately downstream of the catalytic site. The extent of pausing at the termination position determines the efficiency of termination (Gusarov and Nudler, 1999; 2001). The purpose of this pause is to provide enough time for the hairpin to fold at the right distance from the catalytic site. The termination hairpin has to fold in the closed confinement of the RNA exit channel to exert its destabilizing effect on the EC (Gusarov and Nudler, 1999; Epshtein et al., 2007), or, according to the shearing model, to pull RNA from the catalytic site (Larson et al., 2008). In either case, the pause widens the window of opportunity for the hairpin to overcome the energy barrier associated with EC destabilization. Backtracking for one or two nucleotides in this case is sufficient to pause RNAP at the termination point without interfering with hairpin nucleation. Indeed, suppressing backtracking at the termination point either by altering its sequence (Gusarov and Nudler, 1999) or by making the EC less prone to backtracking via RNAP mutations or cooperation (Epshtein and Nudler, 2003; Bar-Nahum et al., 2005) inhibits termination.
T-stretches are crucial elements of the termination process not only in bacteria, but also in eukaryotic cells. Eukaryotic RNAP III and archeal RNAP pause and terminate transcription at T-stretches in the absence of additional factors or apparent RNA secondary structures (Cozzarelli et al, 1983; Campbell and Setzer, 1992; Matsuzaki et al, 1994; Santangelo and Reeve, 2006; Hirtreiter et al., 2010). Although the actual mechanism of RNA release by either of these RNAPs has not been established, it is likely that T-stretch-associated backtracking is a component of this process.
Backtracking may also be involved in RNAP I termination, which requires a site-specific DNA-binding protein (TTF-I) that acts as a roadblock, and a T-rich release element just upstream of the roadblocking site (Lang et al, 1994; Evers and Grummt, 1995). Termination occurs ~10–12 bp promoter-proximal of the TTF-I site, suggesting extensive backtracking by the halted EC. Indeed, congruent with the discussion above, the roadblock is expected to trigger backtracking, especially within the T-rich sequence, causing the roadblocked EC to become subject to dissociation effected by a specialized release factor (Tschochner and Milkereit, 1997; Jansa and Grummt, 1999).
Although the precise mechanism of RNAP II termination remains to be elucidated, it has been well established that pausing over the termination region is required to ensure that it occurs (Birse et al, 1997; Dye and Proudfoot, 2001). These pauses are most likely the result of backtracking, because they were associated with transcript cleavage (Dye and Proudfoot, 2001).
Thus, backtracking appears to be an integral part of the termination process of all cellular RNAPs. An atypical conformation of backtracked ECs, in which the RNA:DNA hybrid is shortened and distorted (discussed below), may facilitate EC destabilization and be a prerequisite for most termination events.
All dividing cells must endure frequent collisions between replication and transcription complexes, which occupy the same DNA track and function at the same time. This is particularly true of bacteria, where the rate of replisome propagation is about twenty times faster than that of RNAP. Since most bacterial active and essential genes tend to be organized co-directionally with replication, co-directional collisions should be more frequent than head-on collisions. Recent evidence demonstrates that DNA damage resulting from such collisions depends on RNAP backtracking (Dutta et al., 2011).
The majority of ECs are stable protein-DNA complexes that must be dislodged by the replisome, regardless of their directionality. The structural organization of the EC (Kettenberger et al., 2003; Vassylyev et al., 2007a) eliminates any conceivable mechanism of replication that does not involve EC dissociation. Indeed, in vitro studies show that the replisome kicks off bacterial RNAP by approaching it co-directionally or head-on (Pomerantz and O'Donnell, 2008; 2010). In the co-directional configuration, bypassing of the EC by the replisome seems to occur effortlessly, whereas the head-on configuration was associated with replisome stalling.
However, co-directional collisions appear to be less benign than traditionally thought. A recent genome-wide analysis of transcription-replication collisions in exponentially growing Bacillus subtilis revealed that co-directional collisions at highly expressed ribosomal genes lead to the disruption and restart of replication (Merrikh et al., 2011). In E. coli, chromosomal double strand breaks (DSBs) are the genome-damaging consequences of co-directional conflicts, which, however, occur only upon collisions with backtracked ECs (Dutta et al., 2011). Normally, such collisions are avoided by anti-backtracking mechanisms that involve active ribosomes and transcript cleavage factors (Proshkin et al., 2010; Dutta et al., 2011) (Figure 2). Translating ribosomes play the primary role in preserving the integrity of protein-coding genes. If translation is compromised (e.g. by antibiotics), anti-backtracking factors (GreA and GreB) as well as transcription termination factors Rho and Mfd, become essential in preventing collision-related DSBs (Dutta et al., 2011; Washburn and Gottesman, 2011). For genes that encode stable RNAs, the anti-termination complex, which is built around ribosomal protein S4 and involves transcription elongation factors NusA, NusB, NusG, and NusE (S10), is likely to function as the principal anti-backtracking factor (Dutta et al., 2011). In addition, the efficient cooperation between multiple RNAPs at those genes (discussed above) diminishes potentially harmful backtracking.
An attractive model that explains DSBs resulting from co-directional collisions originates from in vitro observations that synthesis of the leading DNA strand can be interrupted (Figure 2); once RNAP is displaced, DNAP (pol III) can “jump” to the 3′ end of the nascent transcript and use it as a primer (Pomerantz and O'Donnell, 2008). This jumping by DNAP leads to a single-strand DNA break (SSB) that must be repaired in vivo before the next round of replication converts it to a DSB (Figure 2). Apparently, the DNA repair mechanism is not robust enough to fix such SSBs under conditions of excessive backtracking. According to the model, rapid re-annealing of RNA from displaced backtracked RNAPs generates extended R-loops (RNA:DNA hybrids) because backtracked RNAPs carry longer segments of RNA that are available for re-annealing (the extruded 3′ portion) (Figure 2). Such extended hybrids or R-loops provide accessible 3′-OH termini that could serve as primers for DNA polIII (Figure 2). In contrast, active ECs form hybrids of only ~8 bp (Nudler et al., 1997), which are unstable and cannot survive without RNAP (Figure 3) and, therefore, do not support discontinuous replication. In support of the model, it has been shown that high levels of hybrid-specific RNAse H, as well as RNAP mutations (β H1244Q) that diminish backtracking, eliminate DSBs associated with co-directional collisions in vivo (Dutta et al., 2011).
Because RNAP backtracking provides a mechanistic link between protein synthesis and genome instability (DSBs), it has several important implications for bacterial adaptation and evolution. The ribosome is the principal sensor of metabolic fluctuations and stress. Starvation, proteotoxic challenges, and various antibiotics reduce or eliminate protein synthesis, thereby increasing the probability of RNAP backtracking and formation of DSBs. These same adverse conditions activate stress-induced mutagenesis, which depends on the error-prone DSB repair process (Galhardo et al., 2007) that, in turn, accelerates adaptation to environmental changes, such as acquisition of antibiotic resistance (Galhardo et al., 2007). Thus, RNAP backtracking may contribute to stress-driven bacterial evolution. Indeed, mutation and recombination rates of Gre deficient (backtracking-prone) cells are higher than that of wild type cells (Dipak et al., 2011; Poteete, 2011), whereas survival of such cells depends of the SOS response and error-prone DSBs repair (Dipak et al., 2011). Moreover, as discussed above, the likelihood of RNAP backtracking increases in direct proportion to the frequency of rare codons (codon usage program), which modulate the rate of ribosome movement (Proshkin et al., 2010). This may explain why the mutation frequency is considerably higher for rare codons than for common codons (Alff-Steinberger, 2000) and predicts that many mutational hot spots are associated with local ribosome pausing.
Although protein synthesis is not apparently linked to the anti-backtracking mechanism in eukaryotes, some of the numerous RNA-binding proteins that travel with RNAP II and participate in RNA processing and transport may function to suppress backtracking. Other anti-backtracking mechanisms are clearly important. In contrast to bacterial transcript cleavage factor GreB, which is dispensable under non-stress conditions, its eukaryotic analog, TFIIS, is essential for cell viability (Sigurdsson et al., 2010). This is not surprising, considering that backtracking occurs at numerous positions within transcribed sequences of any given gene (Churchman and Weissman, 2011). Thus, co-directional collisions with backtracked RNAP in eukaryotes are also inevitable. Considering the high evolutionary conservation between bacterial and eukaryotic replisomes and between cellular RNAPs, there is little doubt that such collisions would result in DSBs via the same R-loop-dependent mechanism (Figure 2), and constitute one of the major sources of genome instability. Recent evidence implicating R-loops in genome instability in human and yeast cells support this notion (Helmrich et al., 2011; Wahba et al., 2011).
High-resolution atomic structures of backtracked ECs have become available for yeast RNAP II (Wang et al., 2009, Cheung and Cramer, 2011). These structures differ substantially from those of the on-pathway ECs (Kettenberger et al., 2004; Wang et al., 2006, Brueckner and Cramer, 2008) offering some insights into the basis of the unconventional biochemical and biophysical properties of the backtracked/arrested complexes (Figure 3).
As expected, the 3′ end of the nascent RNA in the backtracked EC appears to be threaded into the secondary channel (funnel) (Cheung and Cramer, 2011). Although this characteristic of the backtracked EC might appear trivial in light of past biochemical data, other structural features of this complex were not anticipated.
An EC backtracked by 9 nt has an unexpectedly short DNA-RNA hybrid (6 bp instead of 8-9 bp) that is tilted toward the bridge helix so that the -1 (relative to the active site) nucleotide base of the template occupies the position normally taken by the +1 nucleotide in the active EC (Cheung and Cramer, 2011) (Figure 3). This, together with displacement of the 3′ end of the RNA from the active site (into the secondary channel), renders the arrested EC incapable of NTP addition. Such a short hybrid would render the non-backtracked EC extremely unstable (Kireeva et al., 2000). Instead, the arrested complex is exceptionally stable and apparently resistant to assisting mechanical force (Forde et al., 2002). The determinants of this exceptional stability lie not in the RNA-DNA hybrid but rather in the arrangement of the nucleic acid scaffold elsewhere.
One factor apparently responsible for stabilization of the arrested EC is the binding of the RNA to the interior of the secondary channel/funnel, where it has been extruded, and with which it normally does not engage (Cheung and Cramer, 2011). In earlier structures of backtracked complexes no assignment could be made for RNA extending beyond position +2, which together with simulations of molecular dynamics led to the prediction of its high mobility within the complex (Wang et al., 2009). In contrast the structure reported by Cheung and Cramer featured a well-defined 9 nt long backtracked RNA that made contacts with one side of the secondary channel (dubbed “backtrack site”) and a trigger loop on the other side. The latter was observed to adopt a new conformation, different from all others reported previously (Kettenberger et al., 2004, Brueckner and Cramer et al., 2008, Wang et al., 2006, 2009) and was designated as “trapped” to indicate its incompatibility with on-pathway elongation (Figure 3). This new, extensive RNA-RNAP interface provides the energy necessary to stabilize this complex (undermined by the distorted and shortened RNA-DNA hybrid), whereas the trigger loop “trapped” in the non-productive conformation by its interactions with the RNA contributes to other impediments to the spontaneous restart of elongation. Forde et al. (2002) invoked the conformational changes in the backtracked/arrested EC to explain why assisting forces that reached the physical limit of single-molecule experiments (the stretching of DNA into a non-native structural form) and twice of the RNAP-DNA binding energy failed to rescue it. Binding of the extruded RNA to the interior of the secondary channel and the “trapped” trigger loop subsequently observed in extensively backtracked EC (Cheung and Cramer, 2011), however, provide an explanation of the “unreasonably large” (Forde et al., 2002) activation barrier required to rescue the arrested RNAP from its backtracked state.
Accumulated information regarding the structure of ECs backtracked by 1 nt or further (9 nt of RNA extruded into the secondary channel/funnel) also reconciles the apparent Brownian ratchet nature of short backtracks that are essentially as irreversible as longer ones. Tyrosine 749 in the second largest (Rpb2) subunit of RNAP II forms stacking interactions with the first backtracked RNA residue (+2) (Wang et al., 2009, Cheung and Cramer, 2011), making backtracking by 1 nt energetically neutral or even favorable (Cheung and Cramer, 2011). Further backtracking, however, would lead to disruption of RNA base stacking by this residue (hence called “gating tyrosine”), creating an activation barrier that can be traversed only under special conditions (e.g. weak hybrid and/or base-stacking interactions), resulting in arrest of the backtracked EC stabilized by a new set of interactions with the extruded RNA (Cheung and Cramer, 2011).
While “gating tyrosine” interactions with RNA appear to limit the extent of the initial backtracking (to 1 nt or more at some sites), its upper limit remains largely unexplored. In their report Cheung and Cramer (2011) provide insight into the natural upper limit for the extent of EC backtracking. Having obtained a highly defined structure of polypyrimidine RNA extruded into the secondary channel/funnel, they observed by modeling that the same backbone with purine bases produced clashes with the RNAP II. This finding not only explains the propensity of ECs to backtrack along pyrimidine-rich RNAs (Hawryluk et al., 2004), but also suggests that the first purine in the nascent RNA encountered by the EC in its retrograde motion along the nucleic acid scaffold may inhibit further backtracking due to steric clashes in the secondary channel.
Altogether, the available X-ray structures of various backtracked complexes help to explain the reasons for the extraordinary stability of the arrested complexes and their resistance to spontaneous reactivation. In the past, the role of the cleavage factors, such as bacterial GreA/B or eukaryotic TFIIS, was seen largely as one of remodeling/reactivating the RNAP active site through the donation of Mg++-coordinating acidic residues to stimulate cleavage of the RNA (Opalka et al., 2003, Laptenko et al., 2003, Sosunov et al., 2003). The emerging realization that the backtracked EC is not equivalent to an active one simply displaced backward along the nucleic acid scaffold, but is a distinct and stable conformational off-pathway state of RNAP, puts the task of its re-activation beyond the mere formation of a new RNA 3′-end in the active site. It appears more likely that even the cleaved extruded portion of the RNA would exhibit a very slow off-rate from its extruded position (due to its extensive interactions with the secondary channel and the trigger loop “trapped” in an inactive conformation), delaying re-activation of the complex. TFIIS soaked into crystals formed by an arrested EC displaces the RNA from its binding site in the secondary channel and restores the trigger loop from the “trapped” to the “locked” conformation (Cheung and Cramer, 2011). A similar interplay between the trigger loop and GreB was proposed based on biochemical studies of backtracked E. coli RNAP (Roghanian et al., 2011). As a result, the re-mobilized RNA can be cleaved off and dissociate, leaving the EC in the state poised for NTP addition.
Backtracking is the fundamental property of all cellular RNAPs that allows the process of transcriptional elongation to be regulated. It provides a means for elongation and termination factors to act on RNAP to control its local transit, overall rate, and accuracy. Backtracking-mediated pausing also plays a major role in transcription termination, co-transcriptional RNA folding and processing. Finally, the association of backtracking with genome instability, at least in bacteria, provides a mechanistic link between growth conditions and cellular adaptation to stress. Backtracking is a remarkable example of how an enzyme's Brownian motion could broadly impact cellular physiology and evolution.
I thank Vladimir Svetlov for his assistance in preparing the manuscript and anonymous reviewers for valuable comments. This work was supported by a grant from the NIH R01 GM58750 (E.N.)
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.