Factor-stimulated intrinsic termination: getting by with a little help from some friends

ABSTRACT

Transcription termination is known to occur via two mechanisms in bacteria, intrinsic termination (also frequently referred to as Rho-independent or factor-independent termination) and Rho-dependent termination. Based primarily on in vitro studies using Escherichia coli RNA polymerase, it was generally assumed that intrinsic termination and Rho-dependent termination are distinct mechanisms and that the signals required for intrinsic termination are present primarily within the nucleic acids. In this review, we detail recent findings from studies in Bacillus subtilis showing that intrinsic termination in this organism is highly stimulated by NusA, NusG, and even Rho. In NusA-stimulated intrinsic termination, NusA facilitates the formation of weak terminator hairpins and compensates for distal U-rich tract interruptions. In NusG-stimulated intrinsic termination, NusG stabilizes a sequence-dependent pause at the point of termination, which extends the time frame for RNA hairpins with weak terminal base pairs to form in either a NusA-stimulated or a NusA-independent fashion. In Rho-stimulated intrinsic termination, Rho prevents the formation of antiterminator-like RNA structures that could otherwise compete with the terminator hairpin. Combined, NusA, NusG, and Rho stimulate approximately 97% of all intrinsic terminators in B. subtilis. Thus, the general view that intrinsic termination is primarily a factor-independent process needs to be revised to account for recent findings. Moreover, the historical distinction between Rho-dependent and intrinsic termination is overly simplistic and needs to be modernized.

Introduction

The bacterial transcription elongation complex (TEC) is highly stable and capable of incorporating several kilobases of nucleotides into a single transcript, such as the 27 kilobase (kb) fla/che gene cluster in Bacillus subtilis [1,2]. A major source of this stability is the suite of biochemical interactions between RNA polymerase (RNAP) and the sugar-phosphate backbone of the RNA:DNA hybrid, with other sources derived from interactions between RNAP and the nascent RNA, and between RNAP and the downstream DNA [3–5]. The disruption of these contacts, leading to deactivation of the TEC, recycling of its components, and transcript release, is an important and highly regulated process. While the global suppression of termination has been found to be a general feature of transcription-coupled repair (TCR) [6], outside of TCR, failure to terminate at the intended transcription end-points greatly alters the expression of downstream genes [7,8]. Even in the absence of a change in the expression of downstream genes, spurious transcription wastes cellular resources. To avoid the numerous consequences of spurious transcription, bacteria evolved several mechanisms to terminate transcription at discrete locations across the genome.

Several transcription factors bind directly to the TEC [9]. By doing so, these factors modulate the transcription activities of the TEC and assist RNAP in recognizing and responding to a variety of regulatory signals [9]. Three of these transcription factors are NusA, NusG and Rho [9–11]. NusA is a highly conserved bacterial transcription elongation factor that is essential in both B. subtilis and Escherichia coli [7,12]. When bound to RNAP, NusA can facilitate the formation of hairpins in the RNA exit channel of RNAP [13–16]. Transcription pausing can be stabilized by the formation of an RNA hairpin in the RNA exit channel [17,18], and NusA is known to stimulate transcription pausing in vitro [14,15] and in vivo [19]. NusG binds to the clamp helices of the β’ subunit of RNAP [20]. Intriguingly, NusG is a pause promoting factor in B. subtilis and a pause suppression factor in E. coli [21–24]. In E. coli, NusG increases transcription processivity by ensuring that the clamp of RNAP remains closed around the DNA binding channel, thus ensuring that RNAP remains in an RNA synthesis competent state [21,25]. Meanwhile, B. subtilis NusG can elicit a sequence-specific pause in transcription by contacting a conserved TTNTTT motif in the non-template DNA (ntDNA) strand within the transcription bubble [22].

Transcription termination is typically described as occurring via two distinct and mutually exclusive mechanisms in bacteria [26–28]. One mechanism, known as Rho-dependent termination, requires the activity of Rho, which is a homohexameric ATPase with RNA translocase and RNA:DNA helicase activities [29–31]. Rho-dependent termination occurs when Rho contacts a cognate RNA element known as the Rho utilization (rut) site and an actively transcribing RNAP [32–34]. The rut site is an unstructured C-rich and G-poor RNA element with regularly interspersed dipyrimidine repeats [29,35]. Recent structural and biochemical studies have shown that E. coli Rho is bound to the TEC, with the C-terminal domains of both NusA and NusG interacting with Rho in the TEC [10,11]. Upon making contact with a rut site, Rho induces an allosteric change in the structure of RNAP such that it enters into a catalytically inactive state, wherein the RNAP β′ clamp is opened and the RNA is dislodged from the active site [10,11,34]. A comprehensive review on this revised model of Rho-dependent termination was published recently [36]. The second termination mechanism, known as intrinsic termination (also frequently referred to as Rho-independent or factor-independent termination), has historically been described as relying primarily on sequence and structural features of the nascent transcript [37,38]. During intrinsic termination, the formation of an RNA hairpin within the RNA exit channel of RNAP induces the melting of a U:A-rich RNA:DNA hybrid [37,38]. Completion of the RNA hairpin promotes transcript release via hybrid shearing, hyper-translocation, or an allosteric model [39–41]. During hybrid shearing, completion of the final base pairs of the RNA hairpin leads to the melting of the upstream portion of the RNA:DNA hybrid [39]. During hyper-translocation, the formation of the RNA hairpin leads to forward translocation of RNAP by several nucleotides without concomitant nucleotide incorporation into the nascent RNA [40]. In the allosteric model, the hairpin invades the RNAP main channel, which causes hybrid melting and other structural rearrangements of RNAP [41]. In each of these intrinsic termination mechanisms, the TEC is sufficiently destabilized via shortening of the RNA:DNA hybrid (the major determinant of TEC stability) such that the TEC components dissociate from one another [3,39–41].

Prior to some recent studies outlined in this review, there were no documented examples of transcription factors stimulating intrinsic termination in vivo. Moreover, intrinsic termination and Rho-dependent termination were considered to be distinct and mutually exclusive processes until recently [26–28,42]. As such, intrinsic termination is typically referred to as Rho-independent termination in virtually all life science textbooks [43]. Recent studies have shown that intrinsic termination in B. subtilis is primarily a factor-mediated process in vivo, with over half of all intrinsic terminators requiring NusA alone, and ~97% of all intrinsic terminators requiring some combination of NusA, NusG, and Rho to terminate efficiently [42].

Term-seq and identification of NusA-stimulated intrinsic termination in vivo

By binding adjacent to the RNA exit channel of RNAP, NusA is able to stimulate the formation of RNA hairpins in the RNA exit channel [13]. This activity of NusA is due to both an allosteric effect of NusA on the RNA exit channel of RNAP and a biophysical contribution of NusA to the chemical environment immediately outside of the RNA exit channel [13,18]. More specifically, the NTD of NusA widens the RNA exit channel so that it can more readily accommodate three strands of RNA (an RNA duplex and the additional strand upstream of the duplex), and also provides positively charged residues that guide the folding of RNA back into the RNA exit channel [13]. For several decades, NusA has been known to cause an increase in the termination efficiency of a limited number of intrinsic terminators in vitro. These intrinsic terminators include the coliphage λ tR2 terminator [16], the E. coli rrnB terminator (T1) [44], the E. coli trp terminator (trp t) [45], and the B. subtilis trp leader terminator [15]. Despite the possibility that NusA stimulated intrinsic termination on a broader scale in vivo, there were several issues that prevented the examination of the in vivo effect of NusA in this process. First, NusA is an essential protein in both E. coli and B. subtilis [12,46]. While it was found that the removal of many horizontally acquired genes in E. coli rendered NusA non-essential [12], the highly synthetic nature of the resulting strain made it a poor model for studying the function of NusA in vivo. Second, several crucial RNA sequencing technologies that allow identification of the in vivo locations of transcript release had not been developed.

Two converging scientific developments overcame these barriers, which allowed for a genome-wide examination of the impact of NusA on intrinsic termination. The first development was the recombineering of a B. subtilis NusA-depletion strain, in which NusA was solely produced exogenously via an IPTG-inducible promoter [7]. Growing this strain in the presence of IPTG resulted in near WT levels of NusA, while growing this strain in the absence of IPTG led to depletion of NusA to about 2% of wild-type levels. The second development was the invention of Term-seq, an RNA-seq-based functional genomic assay that allows for the identification and quantification of all 3’ ends across a transcriptome [7,47]. In Term-seq, the 3’ ends of transcripts obtained from a bacterial culture are ligated to a unique RNA oligonucleotide, thus preserving each authentic 3’ end. The resulting ligation products are then subjected to short read sequencing on the Illumina platform, and the precise 3’ nucleotide that was ligated to the custom RNA oligonucleotide can be identified computationally (Figure 1a). All intrinsic terminators contain an RNA hairpin followed by a U-rich tract ending near or at the point of termination (POT). The in vivo termination efficiency of an intrinsic terminator is calculated by comparing the RNA-seq coverage upstream and downstream of the POT. A 3’ end identified by Term-seq can be the site of transcript release due to transcription termination, an intermediate RNA decay product, or the site of a pause in transcription [7]. By first identifying all 3’ ends that contain the core intrinsic terminator modules (hairpin, U-rich tract) encoded in the upstream sequence, and then thresholding this pool of potential intrinsic terminators by termination efficiency, one can obtain the comprehensive list of intrinsic terminators that function in a particular biological condition from a Term-seq dataset [7].

Figure 1.

Term-seq and a model of NusA-stimulated intrinsic termination. a. Schematic of Term-seq. The protocol starts with growing a bacterial culture to a certain cell density, after which total RNA (green) is extracted and then depleted of rRNA. A unique oligo (red) is ligated to the 3’ end of each isolated transcript to preserve the authentic 3’ end without altering the natural biological sequence (green). These RNAs are then subjected to short-read sequencing on the illumina platform. The sequencing reads provide both RNA-seq and 3’ end information. b. A model of NusA-stimulated intrinsic termination. In this model, NusA (yellow) is bound to RNAP (light blue) during transcription elongation. The DNA, RNA-DNA hybrid, and the nascent RNA is shown. NusA assists intrinsic termination of terminators with bulges in the hairpin (red portion of hairpin) and distal U-rich tract interruptions (red base pairs in the RNA:DNA hybrid).

Employing this approach to RNA extracted from both WT and NusA-depleted strains of B. subtilis made it possible to determine how the loss of NusA impacted the strength of intrinsic termination in vivo on a genome-wide scale [7]. This study identified several hundred NusA-stimulated intrinsic terminators, thereby demonstrating that NusA functions as an intrinsic termination factor in vivo. It was found that this pool of intrinsic terminators tended to have weak hairpins with bulges and distal U-rich tract interruptions (Figure 1b). These observations were recapitulated in vitro, where it was found that inserting a bulge into model intrinsic terminator hairpin stems, or non-U residues into the distal portion of U-rich tracts, increased the NusA dependence of these intrinsic terminators. Moreover, by analyzing the RNA-seq coverage pattern downstream of NusA-stimulated intrinsic terminators, it was found that read through upon the loss of NusA led to increased expression of regulators involved in several important processes, including DNA replication, nucleotide metabolism, and amino acid metabolism. Taken together, this study demonstrated that NusA is a vital intrinsic termination factor in B. subtilis and that loss of NusA leads to widespread misregulation of gene expression [5]. Importantly, this was the first study to firmly argue against the notion that intrinsic termination is primarily a factor-independent process [7].

NusG stimulates a pause at the POT

NusG is composed of an N-terminal (NGN) domain and a C-terminal Kyrpides-Ouzounis-Woese (KOW) domain, which are connected by a disordered flexible linker [48]. The NGN domain binds to the clamp helices of the β’ subunit of RNAP, while the KOW domain is free to interact with other factors such as Rho or the S10 subunit of the ribosome. These interactions play a critical role in transcription polarity and coupling of transcription and translation in E. coli [21,49–52]. E. coli NusG functions as a processivity factor via its ability to ensure that RNAP remains in a nucleotide incorporation competent state [21], while B. subtilis NusG promotes strong pausing of RNAP at ~1600 sites genome-wide [22]. When not promoting pausing, B. subtilis NusG facilitates RNAP processivity by serving as a positive translocation factor, likely via a similar mechanism as E. coli NusG [22,25]. NusG-dependent pause sites contain a consensus TTNTTT motif in the ntDNA strand within the paused transcription bubble, with these critical T residues positioned at −6, −7, −8, −10, and −11 relative to the 3’ end of the nascent RNA at position −1 [22,24]. All intrinsic terminators contain a U-rich tract in the RNA:DNA hybrid, with the most critical U residues being those immediately downstream of the RNA hairpin at positions −6, −7, −8, and −9, relative to the 3’ end at position −1 [38,39]. These U residues correspond to T residues in the ntDNA strand, and thus it was recognized that the NusG-dependent pause motif contained significant overlap with the intrinsic terminator U-rich tract motif. In addition, studies with Mycobacterium bovis proteins showed that NusG stimulates termination at M. bovis intrinsic terminators with suboptimal U-rich tracts in vitro [53]. While NusG was known to stimulate pausing during transcription of the B. subtilis trp leader at the time of the M. bovis NusG study [54], neither the in vivo role nor the mechanism of NusG-dependent pausing, were known.

The first evidence suggesting that NusG participates in intrinsic termination in vivo was obtained concurrently with the finding that NusG stimulates sequence-specific pausing on a genome-wide scale [22]. In this study [22], it was observed that deletion of nusG resulted in an increase in RNET-seq read coverage downstream of intrinsic terminators identified previously in B. subtilis by Term-seq [7]. RNET-seq is an RNA-seq-based method for identifying and footprinting all actively transcribing RNAPs in a genome-wide fashion [22,55]. While a method that quantifies nascent RNA, such as RNET-seq, can be used to identify regions that are actively transcribed, it is more appropriate to utilize a method that quantifies released RNA, such as Term-seq, to gauge the impact of termination. Thus, Term-seq was applied to WT and ΔnusG strains of B. subtilis, and it was determined that NusG serves as another genome-wide intrinsic termination factor, albeit to a lesser extent than NusA [56]. Conducting Term-seq on a strain in which the ΔnusG allele was transformed into the NusA depletion strain showed that these two elongation factors cooperate to stimulate intrinsic termination on a strikingly widespread level, with the median termination efficiency across all intrinsic terminators falling more than 50% upon the loss of these two factors [42,56].

Much like NusA, NusG stimulates termination at intrinsic terminators with weak RNA hairpins and U-rich tract interruptions [56]. More specifically, NusG was found to stimulate intrinsic terminators with consecutive A-U or G-U base pairs at the bottom of the hairpin stem, and U-rich tracts with distal interruptions. While the important pause stimulation activities of NusG and the overlap between the NusG-dependent TTNTTT pause motif and the intrinsic terminator U-rich tract were known, whether NusG truly stimulated intrinsic termination through its pause promoting activities required verification. During NusG-dependent pausing, three amino acid residues (Y77, N81, and T82) within the NusG NGN domain are required for NusG to interact with the ntDNA strand [24]. These residues are H, S, and V, respectively, in the NGN domain of E. coli NusG [24]. Changing these residues to their E. coli counterparts rendered the B. subtilis NGN domain incapable of stimulating pausing, explaining why E. coli NusG does not exhibit pause promoting activities at the TTNTTT motif [24]. Similarly, in vitro transcription termination assays showed that the stimulatory effect of NusG on intrinsic termination is explained solely by the NGN domain of NusG, and changing Y77, N81, and T82 to their E. coli identities rendered the NGN domain of B. subtilis NusG incapable of stimulating intrinsic termination [56]. Moreover, the WT NGN protein was capable of cooperating with NusA to stimulate termination, while the mutant NGN domain was not [56]. Time course assays using a template containing a NusG-stimulated intrinsic terminator provided clear evidence of a NusG-dependent pause at the intrinsic POT. As the hairpin to 3’ end distance for pausing is 11–13 nts, while 7–9 nts are required for intrinsic termination [7,19,22,27], this assay was repeated using a template containing a mutant version of the NusG-stimulated intrinsic terminator in which the terminator hairpin was changed such that the NusG-dependent pause motif was left intact, while the hairpin to 3’ end distance was increased by 2 nt. On this mutant template, NusG still promoted pausing but was no longer capable of stimulating intrinsic termination. Thus, these experiments showed that the NusG NGN domain stimulates intrinsic termination and cooperates with NusA via its ability to contact the ntDNA strand and elicit a sequence-specific pause at the intrinsic terminator POT (Figure 2) [56].

Figure 2.

Model of NusG-stimulated intrinsic termination. In this model, NusG (purple) is bound to RNAP (light blue) during transcription elongation. The DNA, RNA-DNA hybrid, and the nascent RNA is shown. NusG can be seen making contacts with the non-template DNA strand at the positions indicated by the flipped-out bases. NusG assists intrinsic termination of terminators with terminal A-U and/or G-U base pairs (red base pairs in the terminator hairpin) and distal U-rich tract interruptions (red base pairs in RNA:DNA hybrid).

Rho remodels the upstream RNA to insulate the intrinsic terminator hairpin

Discovered in 1969, Rho-dependent termination was the first bacterial transcription termination mechanism to be identified [57]. The most current models of Rho-dependent termination posit that Rho is bound to RNAP during transcription elongation, and that upon contacting a rut site in an appropriate transcriptional context, Rho allosterically modifies RNAP into an elongation incompetent state, which eventually results in transcript release [10,11]. Several reviews that cover Rho-dependent termination have been published, including one that describes this revised mechanism [27,28,36,58].

Akin to the assumed lack of transcription factor involvement during intrinsic termination, Rho-dependent termination and intrinsic termination have been assumed to occur via distinct mechanisms [27,28]. This assumption was based primarily on in vitro experiments using E. coli RNAP. For example, a Rho-dependent terminator (trp t’) was identified ~250-nt downstream of the trp t intrinsic terminator at the end of the E. coli trp operon [45]. These two transcription terminators were found to operate independently, whether encoded in tandem or separately [45].

Several factors have made it difficult to ascertain the effect of Rho on intrinsic termination in vivo. First was the finding that the 3’ ends of transcripts that are terminated by Rho are subject to processive 3’ to 5’ exonucleolytic degradation by PNPase [59]. PNPase activity is impeded by strong RNA structures, and so PNPase degrades these 3’ ends back to a stable RNA hairpin [59–61]. Thus, both Rho-dependent terminators and intrinsic terminators can have highly similar RNA-seq profiles, with sharp coverage drop-offs immediately downstream of RNA hairpins [59]. The difference is that in the case of intrinsic termination, the coverage drop-off is due to transcription termination [7,56], while in the case of Rho-dependent termination, the coverage drop-off is due to the inhibition of PNPase by the RNA hairpin [59]. Due to the unstructured nature of rut sites, it is likely that most Rho-dependent 3’ ends are subjected to this type of degradation, making it exceptionally difficult to identify a comprehensive set of Rho-dependent terminators in vivo [8,62,63]. Second, the RNA hairpin upstream of a Rho-dependent terminator can in some cases be followed by a short U-rich tract. In one case, what was once thought to be an intrinsic terminator, based on the presence of an RNA hairpin followed by a U-rich tract, was actually the end-product of PNPase-mediated degradation that occurred following Rho-dependent termination [64].

Despite these findings, there were several known mechanistic overlaps between Rho-dependent and intrinsic termination. First, both Rho-dependent termination and intrinsic termination occur more readily when the rate of transcription is slow or when RNAP is paused [16,38,65,66]. Moreover, the downstream portion of the intrinsic terminator U-rich tract is known to stimulate a pause in transcription, which precedes transcript release [16,38]. The second is the nature of Nus factor involvement on termination in E. coli and B. subtilis. In B. subtilis, the NGN domain of NusG stimulates intrinsic termination by stabilizing a pause at the end of the U-rich tract [22,24,56], while in E. coli, the KOW domain of NusG serves to coordinate Rho-dependent termination [10,11,50,51]. Third is the observation that both Rho-dependent termination and intrinsic termination utilize somewhat analogous hyper-translocation, hybrid shearing or allosteric approaches to destabilize the TEC, depending on the transcriptional context [34,39–41,67,68].

A major function of Rho is transcription silencing of poorly translated genomic regions in a process known as transcription polarity [12,69]. In transcription polarity, the KOW domain of NusG interacts with Rho to induce transcript release in situations where the KOW domain is not contacting an available S10 subunit of an actively translating ribosome [49,50]. This regulatory switch allows Rho to silence horizontally acquired genes, antisense transcription, and transcription downstream of nonsense mutations [12,69]. In E. coli, the ability of Rho to preferentially silence these regions is contingent upon the coupling of transcription and translation, as the impetus for Rho-NusG interaction is the loss of a pioneering ribosome in the vicinity of the TEC [49,50]. However, in contrast to E. coli, it was recently found that transcription proceeds at a faster rate than translation in B. subtilis [70]. A consequence of the difference in rates between these two processes is that transcription and translation are not extensively coupled in B. subtilis [70].

Both the general lack of coupling between transcription and translation, and the finding that Rho is a component of the TEC, have important implications for the regulation of Rho-dependent termination in B. subtilis [10,11,70]. In addition to these recent findings, experiments were recently performed to reexamine the distinction between Rho-dependent and intrinsic termination. First, Term-seq was conducted on WT and Δrho strains of B. subtilis under the same growth conditions previously used for the ΔnusG and NusA depletion strains [42]. Through a similar set of analyses as described above, 60 intrinsic terminators were identified that decreased in efficiency by ≥25% upon the deletion of rho [42]. These Rho-stimulated intrinsic terminators were found to have particularly weak hairpins and U-rich tracts with interruptions both distal and proximal to the RNA hairpin [42]. The ability of Rho to stimulate several of these terminators was confirmed in vitro (detailed below).

Transcription attenuation mechanisms typically involve partially overlapping antiterminator (AT) and intrinsic terminator hairpins [71–74]. Due to this stretch of shared sequence between a downstream portion of the AT hairpin and an upstream portion of the terminator hairpin, the formation of the AT and the terminator hairpin are mutually exclusive. Moreover, due to its position upstream of the intrinsic terminator hairpin, the AT hairpin has the opportunity to form before the terminator hairpin is fully transcribed, which results in transcriptional readthrough [71–74].

An RNA structure that can bind to flavin mononucleotide (FMN) is encoded in the 5’ leader region of the B. subtilis flavin biosynthetic ribD gene [75,76]. Binding of the RNA aptamer to FMN prevents the formation of an AT, which stimulates intrinsic termination, while the AT is free to form in the absence of FMN [75,76]. This riboswitch mechanism restricts FMN synthesis in B. subtilis to times of FMN starvation [75,76]. A NusG-dependent pause in the ribD leader provides additional time for co-transcriptional binding of FMN to the RNA aptamer, thereby reducing the concentration of FMN necessary to stimulate intrinsic termination [22]. The ribD leader intrinsic terminator was found to be stimulated by Rho in vivo and was thus tested in vitro in the presence and absence of FMN and Rho [42]. It was determined that Rho-stimulated termination at this intrinsic terminator in the absence of FMN but not in its presence. The sfp intrinsic terminator is another Rho-stimulated intrinsic terminator identified in vivo, and it was also shown that Rho stimulated termination of the sfp intrinsic terminator sixfold in vitro [42]. ATs typically only form during transcription attenuation mechanisms in 5’ leader regions [73,74]. In E. coli, AT-like structures that could interfere with intrinsic terminators downstream of protein coding sequences are prevented from forming by the helicase activities of the coupled translating ribosome [70,77]. In B. subtilis, the lack of general coupling between transcription and translation implies the absence of a known factor that constitutively prevents the formation of such AT-like structures [70]. An AT-like structure was identified upstream of the sfp intrinsic terminator hairpin, and deletion of the AT-like sequence increased the termination efficiency to the level observed with the WT template in the presence of Rho [42]. Moreover, Rho was no longer able to strongly stimulate termination of the sfp intrinsic terminator using this deletion template [42]. Furthermore, Rho was unable to stimulate termination of the intrinsic terminator in the ribD leader when the AT was prevented from forming by FMN [42]. Together, these data indicate that Rho prevents the formation of AT-like structures that could otherwise interfere with intrinsic terminators that function downstream of protein coding sequences (Figure 3a). In this model, Rho essentially replaces the ribosome as the primary factor responsible for insulating the intrinsic terminator hairpin.

Figure 3.

Models of Rho-stimulated intrinsic termination and hybrid Rho-dependent termination. a. Model of Rho-stimulated intrinsic termination. Model on the left illustrates transcription when Rho is not contacting RNAP (light blue). In this case an AT-like structure forms outside of the RNA exit channel. Model on the right illustrates transcription when Rho (green) is contacting RNAP. In this case a terminator hairpin forms in the RNA exit channel. The DNA is dark blue and the RNA is dark green. A competing AT-like structure can be seen forming when Rho is not present. The intrinsic terminator hairpin can be seen forming when Rho is present. The red portion of these two hairpins corresponds to the same segment of the nascent transcript. Rho can be seen contacting the nascent transcript upstream of the intrinsic terminator hairpin, thereby preventing the formation of the competing structure. b. Model of hybrid Rho-dependent termination. From left to right. The TEC at an intrinsic terminator. Rho can be seen stimulating an intrinsic terminator. Transcript release is depicted by the upward pointing arrow. In cases where readthrough occurs, a rut site (red) is encoded shortly downstream of the intrinsic terminator sequence. Once transcription has proceeded past the rut site, Rho contacts the rut site and stimulates classical Rho-dependent termination across a broad window, depicted by multiple upward pointing arrows.

Hybrid Rho-dependent termination

In addition to stimulating termination of the sfp intrinsic terminator in vitro, Rho also causes transcript release at several positions downstream [42]. A similar phenomenon was observed in this region in vivo, with a sharp decrease in RNA-seq coverage at the Rho-stimulated intrinsic terminator, while readthrough transcription downstream of the intrinsic terminator was gradually reduced over a more diffuse region [42]. Both of these coverage decreases were abrogated upon deletion of rho. Thus, Rho was found to stimulate both intrinsic termination and non-intrinsic (“classical”) Rho-dependent termination in the same termination window both in vitro and in vivo. A propensity for rut site-like sequences downstream of Rho-stimulated intrinsic terminators was also identified [42]. These results suggest that Rho induces transcript release at classical Rho-dependent terminators by recognizing rut sites encoded downstream of Rho-stimulated intrinsic terminators. Rho facilitating both intrinsic and classical Rho-dependent termination within the same window constitutes a new termination mechanism, which is referred to as hybrid Rho-dependent termination (Figure 3b). Considering the propensity for Rho to reduce spurious antisense transcription [12,69], these results illustrate how Rho utilizes multiple mechanisms to silence spurious transcription in regions that experience heavy antisense traffic.

An updated model of intrinsic termination in B. subtilis

Conducting Term-seq on WT, NusA depletion, ΔnusG, and NusA depletion ΔnusG strains of B. subtilis revealed that NusA and NusG stimulate intrinsic termination both singly and cooperatively [42,56]. In addition, conducting Term-seq on Δrho strains led to the revelation that Rho also functions as an intrinsic termination factor [42]. To better understand the interplay between NusA, NusG, and Rho during intrinsic termination, strains of B. subtilis containing all combinations of NusA depletion, ΔnusG, and Δrho alleles were constructed [42]. In addition, a strain of B. subtilis was generated in which the only source of NusG was the NGN domain [42]. This NGN-only NusG allele was also combined with a Δrho allele [42]. Term-seq was then conducted on all of these strains. Results from this analysis revealed that NusA is the most important intrinsic termination factor in vivo, with NusA stimulating over half of all intrinsic terminators. Moreover, fewer than 3% of all intrinsic terminators were found to function effectively in the absence of NusA, NusG and Rho. Thus, it is apparent that intrinsic termination is primarily a factor-stimulated process in B. subtilis.

To better understand the regulatory interplay between these three elongation factors, the tbcS intrinsic terminator was examined both in vivo and in vitro in the absence and presence of all factors [42]. This terminator was chosen as an example from the pool of Rho-stimulated intrinsic terminators. It was found that NusA has the strongest stimulatory effect on this terminator in vivo. In addition, it was found that NusA cooperates with the NGN domain of NusG to stimulate termination to WT levels. Rho stimulated termination at this intrinsic terminator, but only when full-length NusG was present. Importantly, these results were recapitulated in vitro.

It is now abundantly clear that intrinsic termination in vivo involves the activities of multiple protein factors, all of which cooperate to ensure that this termination mechanism remains a highly efficient process. Thus, we posit the following updated model for intrinsic termination (Figure 4). In this model, the TEC only recognizes and responds to the full suite of intrinsic termination signals when it includes NusA, NusG, and Rho. The NGN domain of NusG stimulates a sequence-specific pause in transcription at the point of intrinsic termination. This pause in transcription provides additional time for NusA and/or Rho to directly facilitate the formation of a terminator hairpin and/or insulate the intrinsic terminator hairpin by preventing the formation of competing AT-like structures. The pause in transcription may also provide additional time for the KOW domain of NusG to recruit Rho to the TEC. From this point of view, the core intrinsic terminator sequence modules (hairpin and U-rich tract) are only partial, albeit vital, components of this termination mechanism. Transcription elongation factors are needed to assist intrinsic terminators with suboptimal hairpins and U-rich tracts, with factor-independent intrinsic terminator being the exception rather than the rule [42]. In cases where intrinsic terminator readthrough occurs, Rho stimulates classical Rho-dependent termination downstream.

Figure 4.

A holistic model of intrinsic termination. Model on the left illustrates transcription of the intrinsic terminator sequence when no transcription factors are contacting RNAP (light blue). Model on the right illustrates transcription of the intrinsic terminator sequence when NusA (yellow), NusG (purple), and Rho (green) are contacting RNAP. The DNA is dark blue and the RNA is dark green. A competing AT-like structure can be seen forming when transcription factors are absent. The intrinsic terminator hairpin can be seen forming when transcription factors are present. The red portion of these two hairpin strands corresponds to the same segment of the nascent transcript. Rho can be seen contacting the nascent transcript upstream of the intrinsic terminator hairpin, thereby preventing the formation of the competing AT-like structure. NusG can be seen contacting Rho. While not shown explicitly here, it should be noted that NusA has also been reported to physically contact Rho within the E. coli TEC [10,11]. In addition, NusG can be seen making contacts with the non-template DNA strand at the positions indicated by the flipped-out bases. Combined, NusA, NusG, and Rho assist intrinsic termination of terminators with terminal A-U and/or G-U base pairs (red base pairs in the terminator hairpin), distal U-rich tract interruptions (red base pairs in RNA:DNA hybrid), and bulges in the hairpin stem.

Perspectives

Our current understanding of intrinsic termination in B. subtilis indicates that there is a substantial involvement of transcription elongation factors [42,56]. As such, attention needs to be directed toward how we refer to this class of terminators. We prefer a nomenclature in which all terminators that rely on a hairpin and U-rich tract continue to be known as “intrinsic terminators”, preceded by a designation that defines their factor dependency profiles. For example, only about 3% of all intrinsic terminators were truly “factor-independent” [42], and thus this minor subpopulation of terminators will be referred to as “factor-independent intrinsic terminators”. Other terminators will be referred to as NusA-stimulated intrinsic terminators, NusG-stimulated intrinsic terminators, and Rho-stimulated intrinsic terminators. Terminators that rely on both NusA and NusG will be called NusA/NusG-stimulated intrinsic terminators and so on. In addition, the distinction between classical Rho-dependent termination and intrinsic termination needs to be deemphasized, as it is clear that Rho stimulates both termination mechanisms in B. subtilis, at times even within the same window. This dual mechanism is called hybrid Rho-dependent termination.

Studies have shown that pause promotion may be a common function of NusG [22,56], but this possibility needs to be substantiated for organisms other than B. subtilis in vivo. Furthermore, Term-seq experiments are required to confirm whether M. bovis NusG can stimulate intrinsic termination in vivo, and assuming that it can, whether the NusG-stimulated intrinsic termination mechanism is conserved between B. subtilis and M. bovis [53,56]. NusA stimulates pausing and intrinsic termination of both E. coli and B. subtilis RNAP in vitro [15,16,19,44,45]. Although NusA-stimulated pausing and NusA-stimulated intrinsic termination have been verified for B. subtilis in vivo [7,19], examination of an analogous NusA depletion strain will be required before these NusA functions can be explored in E. coli. A lack of coupling between transcription and translation may be relatively common [70]. In addition, early studies of Rho found this factor to stimulate the E. coli trp leader intrinsic terminator [78,79], although the mechanism behind this stimulation was never fully investigated. Thus, it is conceivable that Rho stimulates intrinsic termination across a wide range of bacteria. Moreover, the finding that intrinsic termination is largely a factor-stimulated process may be widely conserved as well.

Considering the diverse environments that bacteria populate and the massive span of evolutionary time that separate these organisms, the true scope of regulatory diversity is likely to be immense. As such, perhaps, it is counterproductive to assume that the regulatory strategies utilized by a few model bacterial species are utilized by all bacteria. Moving forward, our research community should make a collective effort to encourage the development of diverse model organisms.

Funding Statement

This work was supported by the National Institutes of Health grant GM098399 to PB.

Disclosure statement

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