A processive riboantiterminator seeks a switch to make biofilms

Summary

Since the discovery of the first signal-sensing RNA structure in Bacillus subtilis, the list of cis-acting riboregulators has grown dramatically. Riboswitches fold into elaborate structures and respond to binding of small metabolites by altering the folding pattern of the surrounding transcript, thereby altering the gene expression program. Riboswitches that use short-range mechanisms to control transcription attenuation and translation initiation and mediate mRNA cleavage have been characterized in many Gram-positive bacteria. Their action typically relies on alternative RNA structures that are differentially stabilized by the ligand binding. In this issue of Molecular Microbiology, Irnov and Winkler describe a novel B. subtilis riboregulator called EAR that shares structural complexity with riboswitches but possesses a unique mechanism of action. EAR increases expression of exopolysaccharide genes and biofilm formation, and appears to act as a processive, long-range antiterminator, the first such example outside of Escherichia coli. While it is unclear whether EAR senses a biofilm-inducing signal, the results suggest that its action depends on yet unidentified auxiliary factors. Interestingly, efficient capsule biogenesis in E. coli and Bacteroides fragilis also depends on processive antiterminators but utilizes the protein-based mechanisms instead.

Multisubunit RNA polymerases (RNAPs) are obligatory processive enzymes that must remain bound to the nascent RNA chain throughout the entire course of its synthesis, for as many as 10⁴ and 10⁶ nucleotide addition steps in bacteria and mammals, respectively. This is quite a daunting task, since movement of RNAP along the DNA template is hindered by many roadblocks: DNA-bound proteins, DNA lesions, errors in the nascent RNA, and nucleic acid signals that induce pausing and termination. Efficient transcription of all genomes from bacteriophages to humans relies on accessory factors that allow RNAP to overcome these barriers (Shilatifard et al., 2003, Weisberg & Gottesman, 1999). Among these factors are processive antiterminators that modify the enzyme into a “productive” state that bypasses consecutive terminators located a few kbs from the modification site, becomes resistant to pause signals, and transcribes at a faster rate. In bacteria, RNAP modification is required for transcription of untranslated or long operons, such as the Escherichia coli rrn (Condon et al., 1995, Greive et al., 2005), rfb (Belogurov et al., 2009) and kps operons (Stevens et al., 1997), phage HK022 early (King et al., 1996) and phage λ early (Rees et al., 1997) and late genes (Grayhack & Roberts, 1982).

Three types of processive mechanisms have been described for the E. coli RNAP (Fig. 1). In the first, RNA-only (King et al., 1996) mechanism (Fig. 1B), a 65-long put RNA structure from HK022 phage directly interacts with RNAP. In the second, hybrid mechanism (Fig. 1C), a small RNA region composed of box A and nut hairpin (box B) recruits λ N protein and at least four host Nus proteins to RNAP transcribing λ early genes (Mason et al., 1992), including Q, the antiterminator for the λ late operon. An analogous mechanism operates in the E. coli rrn operon (Condon et al., 1995). Finally, the protein-only mechanisms (Fig. 1D) utilize DNA-binding antiterminators λ Q and E. coli RfaH that do not require RNA elements for the RNAP modification. In all three cases, the long-range modification requires stable association of the accessory factor(s) with RNAP throughout elongation (Belogurov et al., 2010, Rees et al., 1996, Roberts et al., 1998, Sen et al., 2001). Despite proposed mechanisms of antiterminator action (Belogurov et al., 2010, Gusarov & Nudler, 2001, Komissarova et al., 2008, Rees et al., 1997, Santangelo et al., 2003), the detailed molecular picture of changes in RNAP that accompany its modification remains unclear in each case.

Figure 1.

Processive antitermination. An operon is depicted with a start (a bent arrow) and the end (the finish line shown as a chequered flag). The depth of the gene colour indicates the level of its expression and the transparency of RNAP (shown as a grey oval), its occupancy along the operon.

A. Transcription of long and poorly (or not at all) translated bacterial operons is limited by the presence of intrinsic and Rho-dependent termination signals (shown as stop signs). A fraction of RNAP molecules that initiate transcription at the promoter dissociates at each of these signals, releasing prematurely terminated RNA chains (red lines). Only a small fraction of RNAP reaches the end of the operon, making the full-length transcript (a black line). As a result, the expression levels and RNAP occupancy range from very high at the beginning of an operon to barely detectable near the end.

B. Antitermination by put RNA structure is independent of accessory proteins but requires stable binding of put RNA to the enzyme (Sen et al., 2001). Transcript release is inhibited at each of the consecutive terminators, dramatically increasing the yield of the full-length mRNA. For simplicity, only this RNA species is shown; however, termination is typically not completely abolished (as shown by a partial decrease in RNAP occupancy).

C. Antitermination by λ N requires assembly of a large ribonucleoprotein complex from two RNA elements, box A (cyan) and box B (nut RNA, dark green). λ N directly binds to the nut hairpin and allows RNAP to read through a single terminator when present alone; at high concentration, N can act even without the nut site (Rees et al., 1996). However, the long-lived effect requires several host proteins (NusA, NusB NusE, and NusG) to stabilize the antitermination complex through a network of interactions (Mason et al., 1992).

D. Recruitment of λ Q to RNAP transcribing λ late genes is mediated by the qut DNA element, which overlaps the promoter and directly binds to Q, and a promoter-proximal pause, which is induced by interactions of the initiation factor σ with a -10-like promoter element in the transcribed DNA (Roberts et al., 1998). After recruitment, Q turns into an RNAP subunit and modifies it into a processive state; this modification requires NusA.

E. Antitermination by EAR. After it is transcribed, the EAR element (encoded in the intergenic epsB-C region) binds directly to RNAP and allows it to bypass a strong terminator(s) at the beginning of the epsF gene. Top: EAR binding to RNAP (as shown) and/or its ability to modify the enzyme apparently require additional cellular protein (a magenta oval) or an effector (depicted as a star). Bottom: mutations in EAR or the absence of auxiliary factors lead to a dramatic drop in expression of epsF-O genes.

Interestingly, even though termination is elaborately regulated in Gram positive bacteria by a plethora of proteins, ribonucleoprotein complexes and RNAs, including tRNAs and metabolite-sensing riboswitches (Grundy & Henkin, 2004, Serganov, 2009), processive antitermination mechanism has not been reported until now. In this issue of Molecular Microbiology, Irnov and Winkler close this gap by describing a novel cis-acting ~120 nt RNA element (called EAR; Fig. 1E) that increases expression of EPS (ExoPolySaccharide) genes in Bacillus subtilis (BS) at the level of transcription termination.

EAR is located between the epsB and epsC genes; it was identified during a bioinformatics search for conserved motifs in intergenic regions; all known EAR-like motifs are linked to EPS or biofilm genes, share the sequence and likely the secondary structure elements, and are structurally complex. These features are commonly found in riboswitches, immediately suggesting a mechanism in which direct binding of a (unknown) metabolite controls the elongation/termination decision by disfavouring a nearby terminator RNA structure.

To assess the role of the EAR element in biofilm formation, Irnov and Winkler used gene replacement in a biofilm-forming undomesticated BS strain NCIB3610. They showed that EAR deletions and substitutions that disrupt its conserved secondary structure (as confirmed by inline probing of the EAR structure) decreased biofilm formation. Analysis of the effect of EAR on eps operon expression revealed that mutations in EAR did not reduce epsC, D and E mRNA levels but markedly decreased expression of the distal part of the operon. A dramatic drop was observed after the epsE gene by microarray and qRT PCR analyses, suggesting the presence of a strong termination signal. Indeed, several potential intrinsic terminators were identified in this region. When fused to a lacZ reporter downstream of the EAR element, a tandem of two perfect-looking terminator candidates decreased reporter expression in B. subtilis 10- and 18-fold during exponential and stationary growth in a defined medium, respectively, when a mutant was used instead of the wild-type EAR. Furthermore, EAR worked equally well with heterologous E. coli terminators (λ tr2, T7t, and rrnB T1) in vivo. Thus, EAR acts over a distance and might inhibit termination at diverse signals. These results are in a sharp contrast to a canonical mode of a riboswitch action wherein termination at only one adjacent site is affected, and identify EAR as a processive antiterminator.

The next question addressed by Irnov and Winkler was whether EAR can act alone (as put RNA, Fig. 1B) or functions within an antitermination complex (as nut RNA, Fig. 1C). A self-sufficient factor might be able to act in a heterologous host; for example, a trans-acting non-coding RNA from Listeria monocytogenes can function in E. coli (Loh et al., 2009). However, EAR failed to decrease termination in vivo in E. coli, suggesting that its action requires a host cofactor, an effector, or the cognate RNAP (or a combination thereof). Consistent with the requirement of at least one auxiliary factor in addition to the cognate RNAP, Irnov and Winkler report that EAR fails to act in vitro with purified Bacillus RNAP. Further experiments will be necessary to reveal the identity of EAR cofactor(s) and to elucidate the molecular mechanism of its action.

Irnov and Winkler hypothesize that EAR stably interacts with RNAP as it moves into the downstream genes, and remains an integral part of the eps transcript (Fig. 1E). Based on the mechanism of action of other antiterminators, these are reasonable assumptions. Of course, the most intriguing open question is the molecular mechanism of the EAR-mediated modification of RNAP. Weisberg and co-authors proposed that put RNA, which binds near the RNA exit channel, inhibits the formation or action of the terminator hairpins (Komissarova et al., 2008). EAR could function in this way but several alternative mechanisms can be envisioned, from overall stabilization of the transcription complex against dissociation (at a terminator) to conformational changes in RNAP that would block its isomerization into the off-pathway states. While the road to the detailed answer could be long and bumpy, the initial dataset promises many exciting findings along the way.

Phylogenetic analysis of EAR distribution revealed that it widespread in Bacillales and is positionally linked to biofilm and capsular polysaccharide operons, implying that EAR plays a key role in ensuring the coordinated expression of the early and late genes in these long operons. Biofilm formation represents a drastic change in bacterial lifestyle and, thus, might be controlled by the metabolic state of the cell and external triggers. Interestingly, whereas biosynthesis of surface polysaccharides in Gram-negative bacteria also relies on processive antitermination, a single protein acts as a regulator in these cases. E. coli RfaH, a paralogue of the general elongation factor NusG, is recruited to the elongating RNAP at an ops site located in the leader regions of several operons, including K5 capsule kps operon (Stevens et al., 1997), and switches the enzyme into a highly processive state that ignores pauses and some terminators (Artsimovitch & Landick, 2002). The molecular mechanism of RfaH has been characterized in considerable detail. RfaH structure, binding sites on RNAP and DNA, functional regions (Belogurov et al., 2010), and genomic localization (Belogurov et al., 2009) are already known. It is also known that RfaH does not require accessory host factors or RNA structures, and remains stably bound to RNAP until the enzyme reaches the end of the operon (Belogurov et al., 2009). An orthologous mechanism has been reported in Bacteroides fragilis, the predominant species in the human intestinal microflora. B. fragilis synthesizes eight distinct capsular polysaccharides encoded in eight operons ranging from 11 to 24 kb. Efficient expression of these operons depends on RfaH-like proteins called UpxY; these proteins are encoded in the operon they control, and a small protein region confers their specificity for the cognate operon (Chatzidaki-Livanis et al., 2009). While little is known about the molecular mechanism of UpxY proteins, they share some features with RfaH.

Why do capsule operons need help to produce full-length mRNA? A trivial explanation could be that they are unusually long, and that RNAP will likely encounter termination signals along the way. It is also possible, as Irnov and Winkler argue, that capsule synthesis relies on a physiological signal that could be sensed by antiterminators.

Irnov and Winkler's discovery of a processive bacterial riboantiterminator underscores the amazing diversity of gene regulation by noncoding RNAs and brings up the question of the evolutionary history of gene regulation networks. Although the extant riboregulators cannot be unambiguously identified as descendants of the RNA world, these elements fold into sophisticated structures that enable complex regulation (Serganov, 2009) and serve as scaffolds for binding to coenzymes and nucleotides, the ligands that have existed for billions of years. Furthermore, their broad conservation in bacteria and even eukaryotes (Koonin, 2009) suggests an early evolutionary origin. Figure 1 presents a plausible molecular evolution pathway of a processive antitermination mechanism. The ancient, RNA-only mechanism is exemplified by put RNA, which can function in vitro (albeit with reduced efficiency) in the absence of any protein co-factors (Fig. 1B). The intermediate, hybrid mechanism utilizes two RNA elements (Fig. 1C) to mediate the assembly of a multiprotein antitermination complex that modifies RNAP transcribing E. coli rrn and phage λ early genes (Greive et al., 2005); the RNA is required for the optimal activity but a single key protein can inhibit termination when present at high concentrations (Rees et al., 1996). Finally, the protein-only mechanisms (Fig. 1D) do not require any regulatory RNAs (Artsimovitch & Landick, 2002, Roberts et al., 1998). The current data do not allow us to assign EAR to a particular class: it might require either cellular proteins or a small ligand (or both) to mediate antitermination. In the latter case, the EAR element will share the characteristics of ancient, protein-free riboswitches.