Fail-safe transcription termination: Because one is never enough

Abstract

Termination of RNA polymerase II (RNAPII) transcription is a fundamental step of gene expression that involves the release of the nascent transcript and dissociation of RNAPII from the DNA template. As transcription termination is intimately linked to RNA 3′ end processing, termination pathways have a key decisive influence on the fate of the transcribed RNA. Quite remarkably, when reaching the 3′ end of genes, a substantial fraction of RNAPII fail to terminate transcription, requiring the contribution of alternative or “fail-safe” mechanisms of termination to release the polymerase. This point of view covers redundant mechanisms of transcription termination and how they relate to conventional termination models. In particular, we expand on recent findings that propose a reverse torpedo model of termination, in which the 3′5′ exonucleolytic activity of the RNA exosome targets transcription events associated with paused and backtracked RNAPII.

Introduction

Gene expression is the result of a large repertoire of highly complex and regulated processes. An important layer of regulation that takes place during gene expression occurs initially at the transcription level. Eukaryotic genes are transcribed through cycles composed of 3 sequential, but interconnected steps beginning with transcription initiation and followed by elongation and termination phases, respectively. Each of these steps depend on a tremendous number of factors, highlighting the complex nature and the regulatory potential of the entire transcriptional process.¹ Despite a somewhat fair assessment of the players implicated in transcription, much is still unknown about the regulatory mechanisms underlying each phase of the transcription cycle. This is particularly true for transcription termination of RNA polymerase II (RNAPII). Although it is well accepted that termination of RNAPII transcription is functionally associated with RNA 3′ end formation,^2,3 the relationship between these 2 processes remains poorly understood.

Numerous positive benefits have been associated with proper 3′ end formation/transcription termination, including transcriptional directionality,^4,5 splicing efficiency,⁶ and the control of cryptic transcription events.^7,8 The release of RNA polymerase from the DNA template is also of prime importance for genome partitioning. In this regard, inefficient transcription termination can lead to altered expression of neighboring transcription units by causing either transcriptional interference or collision of elongating polymerases, events that might prove particularly problematic in cases of heavily compartmentalized genomes.⁹ Although it should be reasonably fair to assume that transcription termination is an efficient process, a substantial fraction of RNAPII fails to terminate, presumably owing to unproductive RNA 3′ end formation.^10,11 Consequently, several fail-safe termination pathways, which have been mostly described in yeast, have been uncovered, providing cells with alternatives mechanisms to terminate transcription. Here, we review redundant termination pathways and discuss their mechanism of action relative to conventional models of transcription termination.

Conventional RNAPII Transcription Termination Pathways

In the model organism Saccharomyces cerevisiae, 2 major transcription termination pathways exist depending on the class of RNA transcribed.^12,13 For protein-coding genes, current data suggest a model whereby the co-transcriptional transfer of evolutionary conserved mRNA 3′ end processing factors from the C-terminal domain (CTD) of the largest subunit of RNAPII onto the nascent pre-mRNA promotes cleavage at the poly(A) site (PAS).¹⁴ This endonucleolytic cleavage provides a free and uncapped 5′ entry point for the 5′3′ exonuclease Rat1 (Xrn2 in metazoans), which is thought to chase RNAPII and promote its dissociation from the DNA template by a mechanism that remains unclear and under debate.^15-19 This mechanism, referred to as the torpedo model, was indeed challenged by the allosteric model of termination, which posits that loss of elongation factors and/or conformational changes in the RNAPII complex following transcription of poly(A) signals lead to decrease processivity and termination.^20,21 Still, with data supporting both models, and as these 2 modes of termination are not mutually exclusive, transcription termination probably reflects a combination of these 2 mechanisms, as suggested previously.^22,23

Protein-coding transcription units represent only a small fraction of eukaryotic genomes. Yet, high-throughput sequencing indicates that the majority of genomes are transcriptionally active, yielding a large amount of RNAPII-dependent non-coding transcripts.^24,25 In budding yeast, the vast majority of non-coding RNA genes do not rely on the cleavage and polyadenylation apparatus for transcription termination, but instead terminate using a mechanism that requires the activity of the Nrd1-Nab3-Sen1 (NNS) complex. In this mode of termination, the NNS complex interacts with both a properly phosphorylated RNAPII CTD (phospho-Ser⁵) and specific RNA motifs to engage the transcription elongation complex via the Sen1 helicase, which translocate onto the nascent RNA and ultimately catches up with the transcribing polymerase to elicit termination.^26,27 Even though this mode of termination does not appear to rely on RNA cleavage,²⁸ it shows striking similarities to the PAS termination pathway. Accordingly, NNS-dependent termination relies on 1) the recognition of specifically phosphorylated CTD repeats, 2) binding to sequence-specific RNA motifs, and 3) the displacement of RNAPII by a 5′3′ chasing mechanism. Therefore, even if different RNA-specific transcription termination mechanisms exist, they are driven by common features that apply, partially or totally, to most fail-safe termination pathways.

RNAPII fail-safe transcription termination pathways

A first fail-safe termination pathway involves the RNase type III endonuclease, Rnt1, an enzyme that specifically binds and cleaves double-stranded RNA stem-loop structures capped by a conserved NGNN motif.²⁹ Rnt1-dependent RNA cleavage serves multiple functions in the cell, as it creates unprotected 5′ and 3′ entry points for subsequent maturation or degradation events by exonucleases. Rnt1 cleavage is the first step toward termination of RNAPI,^30,31 but its role in RNAPII transcription is mainly thought as a fail-safe mechanism when defects in PAS-dependent termination occur.^10,32 Mechanistically, Rnt1-mediated termination displays similarities to the PAS termination pathway. Accordingly, analogous to PAS recognition and subsequent endonucleolytic cleavage by the pre-mRNA 3′ end processing machinery, specific hairpin recognition and cleavage by Rnt1 provides a substrate for the 5′3′ exonuclease Rat1, triggering termination by the torpedo model. Such a mechanism was also reported in Arabidopsis, where the RNase type III enzyme, DICER-LIKE protein 4 (Dcl4), provides a backup termination pathway at the FCA locus, which is known to harbor a relatively weak polyadenylation site.³³ Transcription termination via cleavage of RNA structures originating from read-through transcription is reminiscent of another fail-safe termination pathway that occurs at some mammalian genes and that relies on a self-cleaving RNA activity located downstream of PAS signals to allow access to the termination factor Xrn2.^15,34

The aforementioned Rnt1 backup mechanism of termination is unlikely to represent the only option available to the cell, as genome-wide analyses of RNAPII detected Rnt1-dependent transcriptional read-through at a minority of genes,²⁹ although this probably represented an underestimate.³⁵ Accordingly, the NNS pathway is also known to function as a redundant mechanism of transcription termination.¹⁰ NNS components are preferentially enriched at the 5′ end of genes where binding of Nrd1 to serine 5-phosphorylated CTD repeats of RNAPII is predominant.^36,37 Yet, the NNS complex is not restricted to promoter-proximal regions, as ChIP, PAR-CLIP, and CRAC data reveal the enrichment of NNS components at 3′ untranslated regions (UTR) of hundreds of protein-coding genes, arguing for an important role in the control of RNAPII termination.^11,38-40 The fraction of these NNS binding events directly implicated in fail-safe transcription termination remains unknown, however, but was shown to occur at the RPL9B and IPP1 genes.^11,41

Conversely, termination by the NNS complex can also precede PAS-dependent termination, as exemplified at snoRNA genes and at a few protein-coding genes,^11,42,43 where this type of terminator arrangement is likely serving regulatory purposes. Accordingly, for mRNA-encoding genes with such termination signal organization, leakiness of NNS-dependent termination-coupled RNA decay will result in mRNA production due to usage of the downstream PAS. In this case, fail-safe transcription termination is not associated with RNA degradation, as for most of Rnt1- or NNS-dependent termination events.^10,44 Such versatility in the types of terminator arrangement allows transcription termination to be highly flexible. As yet, however, the determinants that promote the use of a specific type of termination event versus another at a given gene are not clear and may simply reflect a stochastic pattern.

A Reverse Torpedo Model of Transcription Termination

Recently, we identified an unsuspected transcription termination pathway in the yeast Schizosaccharomyces pombe that involves the exosome complex of 3′5′ exonucleases,⁴⁵ a machinery that participates in the processing and degradation of multiple RNA classes.⁴⁶ Notably, depletion of core subunits of the RNA exosome results in the widespread production of 3′-extended transcripts from coding and non-coding genes, which correlates with read-through RNAPII at 3′ end of genes, consistent with defects in transcription termination. Furthermore, cases of chimeric “polycistronic” transcripts and transcriptional interference were detected after RNA exosome depletion. These findings argue for an important role for the RNA exosome in fail-safe transcription termination to halt the progression of RNAPII that cannot be dislodged by a 5′3′ torpedo mechanism due to non-productive 3′ end cleavage.

Because transcription termination by the RNA exosome relies on the 3′5′ exonucleolytic activity of its catalytic subunit, Dis3,⁴⁵ a free single-stranded 3′ end substrate must be available. Notably, our data suggest that the generation of a free RNA 3′ end substrate for the RNA exosome is linked to RNAPII dynamics that occur at the 3′ end of genes. Specifically, RNAPII binding studies in various organisms show that RNAPII tends to accumulate at the 3′ end of genes.^47-50 Such pilling up of RNAPII is thought to occur following passage of PAS signals, where a decrease in the elongation rate and subsequent pausing are believed to favor cleavage site recognition and 3′ end processing.⁵¹ Importantly, we found that RNAPII 3′ end accumulation in S. pombe is not limited to pausing, but is also associated with backtracking events. During backtracking, the catalytic center of RNAPII becomes disengaged from the RNA 3′ end and RNAPII slides backward, causing the 3′ end of the nascent RNA to extrude outward from the polymerase,⁵² providing the free single-stranded RNA 3′ end essential for exosome-dependent transcription termination. A criticism of this model argues that the length of RNA associated with backtracked RNAPII is not sufficient to access the active site of the exosome, which requires ∼31-33 nucleotides to traverse the exosome barrel-like structure and reach the active site of Dis3.^53,54 Indeed, single-nucleotide resolution mapping of RNA 3′ end (NET-seq) associated with budding yeast RNAPII has revealed an average backtracking length of 5-18 nucleotides.⁵⁵ However, this backtracking length was derived from pausing events occurring inside ORFs, and may not be an accurate depiction of the actual backtracking that occurs at the 3′ end of genes. Accordingly, comparison of PAR-CLIP and NET-seq data for RNAPII indicates that NET-seq tends to underrepresent RNAPII signals present at the 3′ end of genes.³⁹ Two distinct routing pathways have been reported for an RNA substrate to reach the active site of Dis3: a shorter direct-access route or a through-core route.⁵⁶ Notably, transcription termination defects were observed when the central channel of the core exosome was physically blocked,⁴⁵ arguing for the utilization of the through-core route. It is also possible that binding of the RNA exosome to the extruded RNA favors successive backtracking events, thereby increasing the length of the RNA substrate. However, given the relatively weak interaction reported between short 3′ ssRNA (<24 nt) and the RNA exosome,⁵⁶ a yet-to-be identified cofactor may be required to feed RNAs associated with backtracked RNAPII into the exosome channel. Exosome-dependent termination may therefore be viewed as a “reverse torpedo” model, as it catches RNA and chases RNAPII in a 3′5′ direction instead of the 5′3′ orientation of the conventional torpedo model.

Genome-wide studies have revealed that transcription elongation by RNAPII is not a uniform process with frequent pausing and backtracking events.^55,57 The nature of the determinants that promote RNAPII pausing, backtracking, and susceptibility to exosome-mediated termination remains to be determined, but may involve the chromatin environment.^55,58 Consistent with this view, analysis of nucleosome occupancy in fission yeast indicates some ordering of nucleosomes with no depletion at transcription termination sites,⁵⁹ a pattern that may provide the framework for backtracking events. In fact, impairing the function of chromatin remodelers negatively affects transcription termination.⁶⁰ Beside nucleosomes, other cases of roadblock-mediated transcriptional pausing have also been reported. In S. pombe, the cohesin complex promotes transcription termination at convergent genes,⁶¹ whereas the DNA-binding protein, Reb1, acts as a fail-safe termination mechanism in case of PAS malfunction in S. cerevisiae.⁶² Interestingly, transcription termination defects at convergent genes are present in exosome-depleted cells,⁴⁵ and Reb1-mediated termination is linked to RNAPII backtracking.⁶² Whether the RNA exosome is associated to cohesin- and Reb1-dependent termination pathways remains unknown, but addressing this question should provide valuable insight in defining the elements that predispose pausing sites to exosome-mediated transcription termination.

In this regard, DNA sequences are an important feature of transcription termination by the exosome. Specifically, when sequences known to promote RNAPII pausing⁶³ were mutated at the 3′ end of a model gene, the RNA exosome could no longer exert transcription termination. This suggest that the genomic environment in the vicinity of the pausing site may act to favor a backtracking event that enables exosome-dependent transcription termination. Consistent with this view, RNAPII backtracking is strongly influenced by the strength of the RNA-DNA hybrid inside the polymerase as well as the flanking DNA-DNA duplexes.^64,65 Transcription termination by the exosome can therefore be considered rather universal, contrasting to other fail-safe termination pathways that rely on the recognition of specific RNA sequence/structural motifs.

Teaming Up for Termination?

Nearly a third of all genes in S. pombe show strong transcription termination defects upon exosome depletion. Whether these genes represent cases of inefficient/leaky 3′ end processing is unknown. However, as this termination pathway acts on backtracked RNAPII paused downstream of poly(A) sites, the exosome has the potential to promote termination whether 3′ end cleavage has occurred or not. When 3′ end cleavage is productive, the RNA exosome and Rat1 may thus cooperate in a “double-torpedo” to degrade free 3′ and 5′ ends, respectively, toward optimal transcription termination. Interestingly, we noted a generalized premature dissociation pattern of RNAPII when S. pombe Tfs1 (TFIIS in humans; Dst1 in S. cerevisiae) was absent. By stimulating a slow intrinsic hydrolyzing activity of RNAPII required to realign the RNA 3′ end in the catalytic center of backtracked RNAPII, Tfs1/TFIIS promotes reactivation of backtracked RNA polymerases.^66,67 Backtracked RNAPII stalled downstream of PAS signals are therefore expected to remain inactive for a longer period in the tfs1Δ mutant, and consequently, represent easier targets for termination via the 5′3′ exonucleolytic activity of Rat1 (in the case of productive cleavage by the 3′ end processing machinery) and/or the 3′5′ exonucleolytic activity of the exosome (for both productive and unproductive cleavage). Consistent with this view, we found that the premature termination phenotype detected in the tfs1Δ mutant was suppressed by depleting the RNA exosome; yet, transcription termination defects were still observed in the tfs1Δ dis3 double mutant compared to wild-type cells. Whether this reflects suboptimal transcription termination due to an impaired “double torpedo” remains to be elucidated.

An emerging view of transcription termination is therefore one of a kinetic competition between the transcription reactivation function of Tfs1 and termination by the RNA exosome, Rat1, and possibly both pathways. It remains intriguing why cells use Tfs1 to resume transcription at the 3′ end of genes, especially in cases of dense genome organization such as in yeast. One possibility is the exploitation of context-specific use of alternative polyadenylation sites. Alternatively, the presence of Tfs1 can ensure removal of RNAPII using downstream termination signals when 3′ end formation is not productive at the proximal PAS and backtracking is insufficient to allow exosome-dependent termination.

Conservation of the “Reverse-torpedo” Termination Pathway

The RNA exosome complex is an evolutionary conserved machinery that participates in the post-transcriptional maturation and turnover of numerous RNAs.⁴⁶ Yet, several studies have reported a close association between the RNA exosome and the transcription process in various model organisms.^68-72 In S. cerevisiae, the RNA exosome has been implicated in termination of several non-coding RNA classes, including snoRNAs and CUTs.^73-75 These RNAs rely on the NNS pathway for transcription termination in a process that is intimately coupled to 3′ end maturation by the RNA exosome.⁴⁴ As recruitment of NNS components at a model CUT is impaired in a RRP6 mutant,^44,74 it is possible that the termination defects noted in S. cerevisiae exosome mutants are the result of NNS deficiency. The contribution and mechanism by which the RNA exosome promotes transcription termination in budding yeast therefore remains elusive.

In metazoans, a significant proportion of genes accumulates RNAPII a few nucleotides downstream of their transcription start site in a process known as promoter-proximal pausing. This paused state is thought to serve as a quality control checkpoint for early RNA processing events. Previous work in Drosophila indicates that paused RNAPII are relatively stable, although a substantial amount of RNAPII undergoes premature termination.⁷⁶ Interestingly, RNAPII stalled at promoter-proximal regions cannot resume elongation efficiently unless TFIIS is present.⁷⁷ Such dependence on TFIIS is indicative of backtracked RNAPII and of a free RNA 3′ end that can potentially trigger exosome-dependent termination. In support of this, RNA exosome subunits are present at active promoters in Drosophila.^68,78 However, depletion of TFIIS does not reduce the level of RNAPII at promoter-proximal regions,^77,79 suggesting a minor contribution for the RNA exosome in promoter-proximal termination. In humans, transcription termination defects were reported at the U2 snRNA gene after depletion of a core subunit of the RNA exosome.⁸⁰ Furthermore, backtracked RNAPII is the proposed entry point for the human exosome during class-switch recombination at transcribed immunoglobulin loci, a region known to induce RNAPII stalling.⁷⁰ Although limited, these data provide some indication of the existence of a “reverse-torpedo” model of termination in higher eukaryotes. As a sign of our lack of understanding of human transcription termination, a recent genome-wide mapping of chromatin-bound RNAPII does not detect a transcription termination defect upon knockdown of the termination factor Xrn2,⁸¹ the human homolog of the Rat1 5′3′ exonuclease. Given the important role of the RNA exosome in S. pombe transcription termination,⁴⁵ future studies on pathways of transcription termination in humans are likely to uncover contributions to the RNA exosome.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

Work on transcription termination is supported by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to F.B.. F.B. is holder of the Canada Research Chair in Quality Control of Gene Expression.