Rho-dependent transcription termination: a revisionist view

ABSTRACT

Rho is a hexameric bacterial RNA helicase, which became a paradigm of factor-dependent transcription termination. The broadly accepted (“textbook”) model posits a series of steps, wherein Rho first binds C-rich Rho utilization (rut) sites on nascent RNA, uses its ATP-dependent translocase activity to catch up with RNA polymerase (RNAP), and either pulls the transcript from the elongation complex or pushes RNAP forward, thus terminating transcription. However, this appealingly simple mechano-chemical model lacks a biological realism and is increasingly at odds with genetic and biochemical data. Here, we summarize recent structural and biochemical studies that have advanced our understanding of molecular details of RNA recognition, termination signaling, and RNAP inactivation in Rho-dependent transcription termination, rebalancing the view in favor of an alternative “allosteric” mechanism. In the revised model, Rho binds RNAP early in elongation assisted by the cofactors NusA and NusG, forming a pre-termination complex (PTC). The formation of PTC allows Rho to continuously sample nascent transcripts for a termination signal, which subsequently traps the elongation complex in an inactive state prior to its dissociation.

Introduction

Nearly all bacteria (with a notable exception of Cyanobacteria, Negativicutes, and Streptococcaceae [1–3]) utilize the transcription termination factor Rho for the purpose of separating transcription units, regulating global gene expression, and preserving genomic integrity [2–11]. Discovered in E. coli by Jeff Roberts in 1969 [12], Rho has become the archetype for factor-dependent transcription termination, complementing the intrinsic (factor-independent) termination mechanism within the emerging paradigm of gene regulation in bacteria [13,14]. Whereas intrinsic terminators are generally viewed as simple binary switches, toggled by the formation of RNA hairpins upstream of oligo-U segments in nascent RNA, Rho-dependent termination appeared to offer broader opportunities for regulation through factor recruitment, substrate and signal availability, etc [4,13,15].

Structurally and biochemically Rho is easily identifiable as an ATP hydrolase/RNA translocase [16–20]. In solution Rho spontaneously assembles into a homo-hexameric ring-like motor that binds RNA via its primary RNA binding site (PBS) and threads RNA from 5ʹ to 3ʹ through its central cavity using the secondary RNA binding site (SBS) and energy derived from ATP hydrolysis [4,17,20–23]. Given the established function of Rho as a termination factor and its apparent activity as an ATP-dependent RNA translocase, it is not at all surprising that the bulk of research was focusing on “how” rather than “if” or “how much” Rho translocase activity fit into the mechanism of transcription termination [19,22–26]. Resulting models of Rho-dependent termination then rapidly converged on the mechanism wherein freely diffusing Rho recognizes highly degenerate Rho-utilization (rut) [27,28] sites within nascent RNA as it emerges from transcribing RNA polymerase (RNAP), translocates along RNA toward RNAP, and disengages RNA from the elongation complex once it catches up with the latter [4,6,29].

Among several Rho translocation models, the most widely accepted one, called the tethered tracking model, postulates that Rho stays engaged to rut-site while threading the rest of nascent RNA in a zipper-like fashion [30], further emphasizing the role of the rut-site in termination signaling and mechanism. This model is supported by the observations that PBS-liganded RNA triggers ring closure in vitro, the active state of Rho translocase activity [23], and by single-molecule trapping experiments [31]. The role assigned to RNAP in this and other translocase-centric models is largely that of a passive substrate being acted upon by a powerful molecular motor (RNA translocase), with obscure or altogether absent input from other transcription factors. The passive role of RNAP in Rho-dependent termination can be plausibly traced to the early experiments demonstrating that Rho could arrest and/or terminate transcription by heterologous (yeast) RNAP (pol2, free from bacterial elongation factors) and generate enough torque to dislodge streptavidin from biotinylated RNA [32–34]. Finally, the paradigm of Rho-dependent termination emerged wherein Rho recognized a constitutive terminator (marked by the rut-sites), motored along the nascent RNA until encountering RNAP and shuffling it off DNA by brute force. The alternative model proposed by this lab and emphasizing allosteric aspects of Rho-dependent transcription termination, gained little to no traction [8].

Traditional models appeared satisfactory until a more nuanced picture emerged of the multitude of functions Rho played in the cell. Far from being limited to terminating transcription at a handful of constitutive terminators, Rho was shown to dynamically respond to a variety of signals, from elongation complexes stalled on DNA damage and other roadblocks to the fluctuation in rates of translation of nascent RNA by trailing ribosomes [2,7,13,35–37]. Additional lines of evidence indicated that efficiency of Rho as the RNA translocase correlated poorly with its ability to terminate transcription, and that rut-sites could trigger Rho-dependent termination in trans [38,39]. This apparent caveat to the prevailing paradigm was further underscored by the growing appreciation of the difficulties presented to the diffusion-limited search by the crowded cellular environment [40–42], where the number of Rho targets – nascent RNAs bound to RNAP – was dwarfed by the number of RNAP-free RNAs. rut sites exhibiting low sequence conservation and best characterized as highly degenerate motifs (and in some species absent altogether) would not be in position to significantly impact search efficiency, or aid in discrimination of nascent vs terminated RNAs [43–47]. Hence, one of the most fundamental questions – that of the mechanism by which Rho discriminates between nascent and already terminated RNAs, or is shielded from the latter, remained unanswered.

By 2020 the study of Rho-dependent termination reached a de facto impasse stage. The prevailing translocase-centric model (with variations) was replete with in vitro data, detailing biochemical and structural framework of Rho ATPase/translocase activities, and rut site recognition, but moving no closer to describing the now widely acknowledged multifaceted cellular functions of Rho-dependent termination [4]. The alternative allosteric model proposed by us in 2010 [8] successfully challenged basic postulates of the translocase-centric model(s), but itself lacked sufficient structurally detailed mechanism. No structural data were available for Rho-RNAP complexes, which by and large were considered superfluous, nonspecific, or unstable. To resolve the impasse, we set out to determine the composition and structure of Rho-RNAP complexes in model-independent fashion; instead, we took advantage of the co-purification data, and in vivo and in vitro chemical cross-linking experiments involving RNAP, Rho, and other general transcription factors. We were able to demonstrate that Rho specifically binds RNAP in vitro, wherein this binding depends on elongation factor NusA, and, to a lesser degree, on NusG. Notably the recruitment of Rho to RNAP did not depend on the presence of RNA [38], the critical component of the translocase-centric model. Independently the labs of Irina Artsimovitch and Markus Wahl arrived at a similar pathway of RNAP-Rho complex assembly, using a different set of genetic and biochemical data [39]. In our work we set to elucidate the structures of distinct Rho-EC complexes featuring RNA of predetermined length; Wahl and colleagues endeavored to capture metastable/ensemble states of Rho-EC formed along the termination pathway. Invariably these pre-termination complexes (PTC) contained RNAP, NusA, NusG, and Rho hexamer, supplemented with DNA-RNA scaffolds [38,39]. Single-molecule cryo-EM interrogation of these on-pathway complexes yielded crucial insights into the structure of intermediate stages leading to transcription termination, discussed in detail below.

Rho engagement with RNA polymerase – PTC formation

Traditional models of Rho-dependent transcription termination postulate that Rho tracks along the nascent RNA toward RNAP and only contact RNAP at the final moment of termination [21,29]. Rho action on archaeal and yeast RNAPs even lead some to infer that Rho did not form specific complexes with its cognate (bacterial) RNAP and that RNA binding was a prerequisite to Rho interacting with the elongation complex (EC) [33,34,48,49]. However, the new cryo-EM structures revealed that, aided by two general elongation factors, NusA and NusG, Rho is tightly bound to EC, forming a specific pre-termination complex (PTC) long before the termination event. Incidentally in vitro the Rho-RNAP-NusA(±NusG) complex can be assembled without nucleic acids [38]; this clearly indicates that nucleic acids do not supply any scaffold or tether for Rho entering PTC. NusG role in Rho-dependent termination has been appreciated for a long time [50–52], albeit without a cohesive and realistic model of its engagement; NusA was widely assumed to act as Rho antagonist and hence was largely dismissed from Rho recruitment models – despite the early work by Chamberlin, Imai, and others indicating that NusA facilitates recruitment of Rho to RNAP and Rho-dependent termination in vitro and in vivo [10,53–55]. Cryo-EM data obtained by us and others rectified the role of Rho co-factors, NusA and NusG, particularly that of NusA, simultaneously as recruitment factors and modulators of Rho at post-recruitment steps [38,39].

The PTC structures reveal that Rho maintains a hexameric open ring conformation (“washer nut”) to interact with RNAP α and β subunits through its N-terminal domains (NTDs) (Figure 1) [38,39]. One subunit of Rho interacts with the α-helical insertion I9 domain of the RNAP β subunit while another makes contacts with a loop region in front of the clamp pincer domain of RNAP β subunit. The third RNAP-binding subunit of Rho interacts with the C-terminal domain (CTD) of RNAP α subunit and simultaneously binds to NusA-NTD. These specific Rho-RNAP interactions are important for termination. Substitutions of RNAP residues on the interface with Rho subunits compromise termination in vitro and in vivo [38]. This functional importance of specific Rho-RNAP interactions argues against the “shearing” mechanism of termination [31], in which RNAP plays a passive role while Rho “shears” the RNA from the RNA-DNA hybrid in the active site by mechanochemical force. Indeed, the shearing model cannot explain the dependence of Rho termination on specific interactions with RNAP that occur long before the termination event [38]. Neither it can explain the ability of a small-molecule tagetitoxin, which specifically binds the trigger loop – a mobile element of the RNAP catalytic center, to completely abolish Rho-dependent termination [8,56]. Furthermore, the shearing model is incompatible with the ability of rut RNA to act in trans [38].

Figure 1.

(a). Overall structure of E. coli Pre-Termination Complex (PTC) with a short nascent transcript and the transcription factors NusA and NusG. The figure highlights: I. Rho uses an open-ring hexamer conformation to interact with RNAP at the rear face; II. NusA at its canonical site, interacting with Rho and RNAP around the RNA exit channel; III. NusG-NTD at its canonical site, interacting with Rho-NTD and RNAP. (b). Same as (a) but rotated as indicated, Rho-interacting RNAP β subunit domains are highlighted

This multi-point Rho engagement with RNAP facilitates efficient PTC formation at the early stage of transcription elongation. The RNAP module of PTC remains in the conformation that observed in active, elongation-competent EC, arguing that binding of Rho in the tracking mode does not introduce significant changes in EC processivity [38,39] (if anything Rho can suppress a subset of pauses [8]).

Rho engagement with RNA in PTC

Rho prefers C-rich, G-poor unstructured sequences, called rut sites, for the initial RNA loading [57]. Biochemical and structural studies have revealed that Rho uses its primary RNA binding site for loading onto rut sites, which lie upstream of termination site and are approximately 60–90 nt long [58]. Each of Rho monomers binds tightly to two pyrimidines, interspaced by a 7–8 nt loop region [17,18]. The termination sites typically follow within 10–20 nt of the rut site [31]. Based on a series of standalone Rho-RNA binary complexes structures, it has been proposed that rut RNA first binds to the PBS of an open-ring Rho and then RNA is guided into the central channel comprised Rho-CTD [17]. The loading of rut RNA isomerizes Rho into a closed-ring state that is competent for ATP hydrolysis and translocation [59] (Figure 2a and b).

Figure 2.

Structures of Rho and PTC structure with a long RNA (PDB: 6XAS) (a). Protomer structure of Rho (subunit b of PDB 5JJI). Rho structure is shown in green surface representation, bound ADP in stick representation, rU7 RNA (SBS ligand) in Orange and PBS ligand RNA in yellow (modeled from PDB:1PV4). (b). A proposed sequential RNA loading by Rho in which rut RNA is first loaded on the primary binding site (PBS; indicated by black dashed circle), then presented to a secondary binding site (SBS; indicated by blue dashed circle) and thus triggers ring closure and ATP-driven translocation. (c) E. coli Pre-Termination Complex structure with a long RNA. The subunits are colored as Figure 1. The yellow dash line shows the possible path of the nascent RNA loading into Rho central channel. According to the PCT structural model, nascent RNA interacts first with SBS in the central channel of Rho and then is presented to PBS

In the PTC, the 5ʹ segment of a partial rut RNA was first seen in the central channel of Rho, rather than in the primary binding site of Rho, indicating that the RNA loading steps may happen in a different order (Figure 2c). The prevailing “textbook” model of Rho-dependent termination postulates that Rho need to load onto rut RNA for the purpose of “catching up” to the moving EC, with the corollary that rut RNA is first recognized by the PBS of Rho, then sequentially interacts with P-, Q- and R- loops of Rho-CTD to trigger ring closure and ATP-driven translocase activity (Figure 2b). However, in the PTC (and, likely, in many EC complexes in vivo), Rho has been already pre-bound to RNAP, which provides two distinct advantages over the diffusive search for rut sites in nascent transcripts: 1) Rho is co-localized with the nascent RNA through its interactions with RNAP and Nus co-factors in the elongation complex; 2) Rho is sequestered away from the bulk of already terminated RNAs. Structural models of PTCs suggest that, upon PTC formation, the nascent RNA emerging from the RNAP exit channel occupies the “secondary” binding site in the Rho central channel first, entering the Rho ring from its C-terminal side. As the RNA is being threaded through the central channel, the rut sequence will be presented to the “primary” binding sites located in Rho-NTD. Once all six Rho-NTDs are loaded with rut RNA, the ring fully closes triggering activation of the ATP-driven translocase. It bears noting that this revised mechanism of RNA loading does not contradict the previous findings derived from the structural models of Rho-RNA binary complexes, as it merely postulates a different order in which these complexes occur along the pathway. Moreover, the revised order of RNA loading explains Rho-dependent termination mediated by rut RNA in trans [38], in which rut RNA was added to trigger the termination of the PTC with a relatively short nascent transcript that is not long enough to reach Rho NTD and to then be loaded to the central channel.

The role of NusA and NusG in Rho-dependent termination

NusG was first identified as part of the λ N anti-termination complex [60] that is capable of accelerating the elongation process by suppressing RNAP backtracking pausing [61–63]. NusG was found to facilitate early termination at several Rho-dependent terminators in vitro [51,64]. It has also been demonstrated that NusG-CTD interacts with Rho while NusG-NTD directly binds RNAP via β’-clamp helices of RNAP [61,65,66]. Thus, it was proposed that NusG, as a cofactor, will enhance the stability of Rho-RNAP interaction. This notion is also supported by the fact that about one-third of Rho-dependent terminators are NusG-dependent in vivo [44,67].

The PTC structures confirm that the NusG-NTD is bound to the upstream face of the EC cleft (Figure 1). The location and orientation are consistent with a cryo-EM structure of E. coli EC-NusG [61]. NusG-NTD bridges across the β and β’ subunits of RNAP on top of the upstream duplex DNA, constraining the path of upstream DNA, similar to the action of the NusG eukaryal ortholog, Spt5 [68]. The density for NusG C-terminal domain (NusG-CTD) was ambiguous in the PTC structures. Inter-protein cross-links between NusG-CTD and RNAP, NusG-CTD and Rho suggest NusG-CTD extends toward the β’ zinc finger domain (ZFD) of RNAP and interacts with one of the Rho subunits NTD. This dual affinity of NusG facilitates the PTC formation early in the transcription cycle without the need for nascent rut RNA. It is also consistent with previous reports that NusG-CTD mediates EC and Rho interaction [69]. Structural model of the Rho-NusG-CTD binary complex proposed that NusG modulates Rho-dependent transcription termination by promoting Rho ring closure [70]. In contrast, in the Rho-EC complexes, Rho remains in the open-ring conformation in presence of NusG [38,39]. In the binary complex, NusG-CTD interacts with Rho-CTD, thereby capturing and stabilizing Rho in a close-ring state on suboptimal rut RNAs [70]. However, as seen in the PTC, Rho binds at the rear face of EC (in direction of transcription) while Rho-NTD oriented toward RNAP and NusG-NTD binds at its canonical site (Figure 1b). This relative orientation of Rho and NusG exclude the possibility that NusG-CTD could reach Rho-CTD, indicating that NusG-CTD and Rho-CTD interaction observed outside of its biologically relevant context (EC) may be superfluous.

NusA is a multi-domain transcription elongation factor that is composed of the N-terminal RNAP-binding domain, three RNA-binding domain (S1, KH2 and KH2) and two C-terminal acidic repeats (AR1 and AR2) [71]. It was observed that the presence of NusA delays the Rho-dependent termination window in in vitro transcription assays [53]. Given the potential competition for RNA between NusA and Rho, as well as the effects of NusA and Rho mutants in tiling microarray, it has been proposed that NusA is a direct competitor of Rho for RNA binding and an antagonist of Rho-dependent termination in E. coli [72].

The bulk of data argues that, rather than being an antagonist, NusA facilitates recruitment of Rho to E.coli RNAP, and stimulates its activity [10,53,54,73]. Indeed, the PTC structures reveal that NusA uses its RNAP-binding NTD to interact with Rho, thereby stabilizing the PTC conformation by providing an additional anchor for Rho binding to the EC (Figure 1). The S1 domain of NusA serves as an extension of the RNAP RNA exit channel, connecting it with one of the Rho ring subunits [38,39]. Together with the β’ ZFD of RNAP and Rho subunit, it forms a positively charged multi-protein path for the nascent RNA, which loads RNA into the central channel of Rho (Figure 2c). In vitro transcription assays confirm that in presence of NusA, the efficiency of Rho-dependent termination increases even though the termination window is being shifted downstream, which indicates that NusA actually facilitates termination by stabilizing PTC and interacts with the nascent transcript before it reaches Rho [38,39]. These biochemical and structural results support our earlier in vivo observations demonstrating the role of NusA as a global potentiator of Rho-dependent termination [10].

Transition from the PTC to the trapped (inactivated) termination complex

Cryo-EM structures of three key states of EC-NusA-NusG-Rho-rut RNA complex illustrate the stepwise transition from the PTC to the termination complex, including rut RNA capture, EC rearrangement, and permanent EC inactivation [8,38,39]. During the termination phase, Rho allosterically inactivates the EC, prior to the complex dissociation, via interactions with RNAP, NusA, upstream DNA and rut RNA, which lead to remodeling of the active site and its substrate [8,38,39]. Here we outline our vision of the pathway [8,38], which is in many ways congruent to the one proposed by Artsimovitch and Wahl labs [39]. The main features these models have in common are Rho recruitment that depends on NusA and NusG, but not on RNA, the widening of the DNA-binding channel, and the rearrangement of the catalytic site as obligatory steps in EC inactivation and eventual dissociation, and also the displacement of NusG NTD from DNA at the late stages of termination pathway. For the detailed enumeration and illustration of the steps in Said et al. version of the allosteric model the reader is directed to the series of supplementary movies [39]. We believe that the essentially congruous allosterism of these termination models and their key structural transitions will hold up to the future research, whereas the exact sequence of events leading from Rho recruitment to the EC inactivation and dissociation will be elucidated through additional biochemical and structural interrogation.

The transition starts with rut RNA capturing, concomitant with a drastic conformation change in the Rho hexamer ring. One subunit of Rho switches from one side of the ring opening to the other in order to “step down” to the ZFD of RNAP β′ subunit and “catch” rut RNA (Figure 3b). The rearrangement of the Rho hexamer ring displaces NusG-NTD from EC, while making contact with the upstream DNA. By simultaneously binding rut RNA and upstream DNA Rho induces conformational changes of the lid and clamp domain of RNAP β′ subunit. More specifically, the β′ clamp rotates away from the RNAP central cleft, widening the nucleic acid channel of RNAP. Meanwhile, the acceptor nucleotide of the template DNA at the RNAP catalytic site is shifted to a paused state (Figure 3c). Next, RNAP loses its grip on the DNA-RNA hybrid with a wider opening clamp. The 3ʹend of nascent RNA is displaced together with the template DNA from the catalytic site of RNAP, thus completely inactivating the EC (Figure 3d). Notably, Rho retains the open-ring conformation during this transition, which indicates no helicase or translocase activity of Rho is required in these steps [39].

Figure 3.

Conformational changes during the transition from the PTC state to the EC inactivation state in Rho-dependent transcription termination (a-d). Lower panels highlight the RNAP β’ clamp movement. The β’ clamp is shown in light blue, DNA in blue, RNA in yellow, the catalytic site of RNAP is indicated with a yellow star. Other subunits are colored as in Figure 1. RNAP catalytic site of the Inhibition state is also shown in the close-up view to highlight the movement of the acceptor nucleotide. In lower panels (c) and (d), the close-up views are rotated 45° relative to the original that shows the overall structures. PDB codes for the corresponding structures are (a) 6XAS; (b) 6Z9R; (c) 6Z9S; (d) 6Z9T

The allosteric model of Rho-dependent transcription termination

In contrast to the traditional “catch-up”/translocase-centric model, up-to-date structural, biochemical, and genetic data are broadly consistent with the alternative “allosteric” model [8], and generally incompatible with “shearing” or forward translocation models. In the updated allosteric model, the actual termination process starts with the PTC formation wherein Rho binds to RNAP, NusA and NusG, prior to contacting the nascent RNA [38,39]. The formation of PTC positions Rho in tracking orientation relative to the RNAP RNA exit channel, thereby allowing it to scan the nascent RNA for termination signals without engaging its translocase activity and ATP hydrolysis. Once termination signal (rut RNA) emerges from RNAP, Rho rearranges its open-ring conformation to capture it. The resulting conformational changes lead to the irreversible inhibition of the EC, manifested by the RNAP clamp opening and the inactive/paused state of acceptor nucleotide in the RNAP catalytic site – a process reminiscent to the allosteric mechanism of hairpin-dependent (intrinsic) termination [74,75]. Rho continues to further widen the clamp concomitant with displacing RNA 3ʹ end from the catalytic site. RNAP gradually loses its grip on RNA-DNA hybrid (one of the major determinants of EC stability), again, similar to our allosteric model of intrinsic termination [74], which in turn leads to the dissociation of the otherwise extremely stable EC. Said et al. noted that their late-stage complex (complex IV) exhibited similarities to the paused bacterial EC and yeast pol II inhibited by α-amanitin [39], namely the opening of the clamp, widening of the DNA-binding channel, and destabilization of the template DNA acceptor nucleotide – the features also associated with the allosteric model of intrinsic termination [74].

Perspectives

The two definitive reports of structural basis for Rho-dependent termination in vitro we discussed here introduced a detailed, physically realistic structural and mechanistic model for the recognition and execution of the classic termination signals (rut sites) in nascent RNA [38,39]. However, Rho impacts a multitude of cellular processes, far beyond constitutive Rho-dependent termination, from operon polarity and termination of untranslated mRNAs to suppressing spurious transcription and RNA loops formation, to regulating gene expression by riboswitches, natural aptamers, and small RNAs [2,6,7,10,11,35–37,76–79]. Whereas our understanding of Rho action in the context of canonical Rho-dependent terminator (RNAP, NusA, NusG, Rho, and rut-containing nascent RNA) gained a solid structural and extensive mechano-chemical framework [8,38,39], these findings have limited bearing on other cellular functions of this factor. We expect that these functions are largely or completely rut-independent, but, instead, involve association with additional protein and RNA co-factors, tailoring Rho activities (translocase, helicase, RNA-binding, EC modulation, inactivation, and dissociation) to specific pathways. Structural and biochemical interrogation of these pathways is predicated upon discovery and functional discrimination of such complexes in vivo, via high resolution structural proteomics and interactomics of Rho. Given the importance of Rho in bacterial physiology and genetics, and the importance of both fundamental and translational bacteriology in medicine and biotechnology [80], the continual progress of the latter would be greatly accelerated by our better understanding of the former.

Funding Statement

This work was supported by the National Institutes of Health [R01 GM126891].

Disclosure statement

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