Control of RNA pol II speed by PNUTS-PP1 and Spt5 dephosphorylation facilitates termination by a “sitting duck torpedo” mechanism

RNA pol II transitions into slow elongation near the poly(A) site

A conserved feature of metazoan genes is that apparent pol II occupancy measured by ChIP or nuclear runon peaks in a broad region downstream of poly(A) sites, the termination zone, prior to release of polymerases from the template (Anamika et al., 2012; Core et al., 2008; Davidson et al., 2014; Glover-Cutter et al., 2008; Lian et al., 2008). High pol II signals in termination zones could reflect a slowdown in elongation resulting in a pile up (Ehrensberger et al., 2013) or alternatively, a conformational change in the EC or the chromatin template (Spies et al., 2009). We set out to investigate this question by monitoring the kinetics of transcription through 3’ flanking regions. In metazoans, unlike yeast, global measurement of transcription speed is possible in 3’ flanking regions because intergenic distances are large and pol II can be “synchronized” at start sites by imposing a reversible block with the inhibitor, DRB (Singh and Padgett, 2009). We performed anti-pol II ChIP-seq at intervals after release from a DRB block, which establishes a wave of pol II that transits through the gene and into the termination zone. To minimize loss of the polymerases we are tracking through the termination zone, they were spared from eviction by inhibiting termination with a dominant active site mutant of Xrn2, D235A (Fong et al., 2015). As expected, following DRB washout, the front of the pol II wave quickly progressed through gene bodies, but when it encountered the poly(A) site a transition occurred, marked by a build-up of pol II density over time that was enhanced by Xrn2 D235A (Fig. 1A, S1A-C). Even many minutes after the pol II wave reached the termination zone it did not extend more than a few kb past the poly(A) site (Fig. 1A, S1A-B). To quantify elongation rates in the gene bodies and termination zones we employed a two-state Hidden Markov Model to identify pol II wave fronts. The median elongation rates of pol II within gene bodies were 2.4 kb/min and 3.0 kb/min in Xrn2WT and Xrn2D235A cells respectively (Fig. 1B). Importantly, downstream of poly(A) sites elongation rates fell to just 0.9 kb/min in both Xrn2WT and Xrn2D235A expressing cells (Fig. 1B).

Elongation rates determined by DRB release could be overestimates because the position of the front edge of the pol II wave is determined by the fastest members of the population. We therefore also employed a complementary approach that follows the back edge of a receding wave of EC’s at intervals after inhibiting transcription initiation with triptolide (Titov et al., 2011). As expected, elongation rates within gene bodies estimated from the position of the back edge of pol II wave (1.5-1.6 kb/min, Fig. S1F) are slightly slower than those based on the front edge after DRB release (Fig. 1). Most importantly, we observed a discrete transition at the poly(A) site where the progress of the wave of receding ECs almost totally stalled and pol II density accumulated to high levels in both Xrn2WT and Xrn2 D235A cells (Fig. 2A, B red arrows, S1D). Tracking of the peak of pol II density downstream of poly(A) sites over time after adding triptolide revealed elongation rates of only ~0.1 kb/min in the Xrn2 mutant (Fig. 2C, D). Notably the 5’ boundary of the pol II pile up in both Xrn2 WT and Xrn2D235A cells that marks where deceleration occurs is positioned within 100 bases downstream of the poly(A) site (Fig. 2C, D, S1E). Together the results in Figures 1 and 2 show that elongation rate is severely reduced in the termination zone, and that this slowdown starts close to the poly(A) cleavage site leading to accumulation of high polymerase density (Ehrensberger et al., 2013), especially if termination is prevented.

poly(A) signal recognition is required to slow down elongation

A hallmark of pol II termination is that it is a poly(A) site-dependent phenomenon (Logan et al., 1987; Whitelaw and Proudfoot, 1986). If the pol II slowdown we observed in 3’ flanking regions is relevant to normal termination, then it should require a functional poly(A) site. To test this prediction, we inhibited poly(A) site recognition in HEK293 cells by inducing expression of the influenza NS1A protein (Fig. S2A). NS1A interferes with 3’ end processing by binding to CPSF30, which recognizes the AAUAAA motif in the pre-mRNA (Chan et al., 2014; Nemeroff et al., 1998; Schönemann et al., 2014). We investigated pol II occupancy at 3’ ends of genes 24 hours after NS1A induction, when there was no detectable effect on cell viability. As expected, NS1A caused widespread inhibition of transcription termination manifested by readthrough transcription far downstream of poly(A) sites (Fig. 3A-C, S2B, E) (Eaton et al., 2018; Heinz et al., 2018). This readthrough transcription was specific to genes with poly(A) sites and was not observed downstream of NEAT1 (Fig. S2C) which encodes poly(A)- transcripts cleaved by RNAseP (Sunwoo et al., 2008). NS1A reduced the pile up of pol II immediately downstream of poly(A) sites at individual genes and metagene analysis showed that this behavior typified a majority of genes throughout the genome (Fig. 3A-C, S2B, E). This effect of NS1A was specific to the peak of pol II at 3’ ends while the promoter-proximal peak of pol II density was relatively unaffected (Fig. S2D). Elongation rate measurements in the presence of NS1A were not possible as DRB appears to interfere with NS1A function. Nevertheless, these experiments demonstrate a genome-wide suppression of pol II accumulation in termination zones by NS1A that is consistent with the model that poly(A) site recognition by CPSF is required for pol II to slowdown.

To test further whether poly(A) site recognition is required for pol II deceleration, we asked whether this phenomenon requires the AAUAAA consensus sequence. We replaced the AAUAAA element in the poly(A) site of the chromosomal PSMG1 gene with a BamHI site (DpA) in HAP1 cells by CRISPR/Cas9 gene editing (Fig. 3D). PSMG1 was selected because it is not essential and has only one annotated poly(A) site. Replacement of the wild type sequence was confirmed by showing that a genomic PCR amplicon flanking the PSMG1 gene 3’ end is completely digested by BamHI (Fig. 3E). Anti pol II ChIP-seq revealed the characteristic pile-up of pol II density just downstream of the poly(A) site in the wild type PSMG1 gene, but in the ΔpA mutant it was shifted approximately 5 kb further downstream to a new location that coincides with the presence of multiple potential cryptic poly(A) sites with AAUAAA sequences (Fig. 3F). We conclude that a poly(A) consensus sequence is required to induce the pol II pile up immediately downstream of PSMG1 and, by extension, that poly(A) site recognition is required for pol II deceleration consistent with a previous in vitro transcription experiment (Nag et al., 2006).

Poly(A) site dependent loss of Spt5 phosphorylation

Our results raise a critical question: how does poly(A) site recognition lead to pol II deceleration? Based on previous work in yeast (Baejen et al., 2017; Parua et al., 2018), we enquired whether Spt5 could be involved in an “allosteric” change in the EC at poly(A) sites. Anti-Spt5 ChIP-seq showed that, like pol II, Spt5 accumulated in 3’ flanking regions when termination was inhibited by Xrn2 D235A. This accumulation is evident on individual genes and metaplots of several thousand genes show it is a general phenomenon (Fig. 4A, B, S3B). We conclude that the pol II-Spt5 interaction persists within slow ECs in the termination zone.

We next assayed phosphorylation of Spt5 by ChIP-seq with antibodies that detect total Spt5 or its phosphorylated isoform specifically (Fig. S3A). Inspection of individual genes (Fig. 4C-E) and metaplots of phospho-Spt5 normalized to total Spt5 ChIP-signals revealed two notable features that applied globally. First, Spt5 phosphorylation increased progressively over the first 2-3kb downstream of promoter-proximal pause sites (Fig. 4F, S4G). Second, Spt5 phosphorylation declined sharply within 100 bp downstream of poly(A) sites (Fig. 4G, S4F). Consistent with a functional connection between Spt5 phosphorylation and elongation rate, we note that the transition to a Spt5 hyperphosphorylated state near the 5’ end of the gene body roughly coincides with where pol II accelerates (Danko et al., 2013; Jonkers et al., 2014), and the return to a hypo-phosphorylated state downstream of the poly(A) site coincides with where it decelerates (Fig. 1, 2).

To ask whether loss of Spt5 phosphorylation downstream of poly(A) sites is dependent on termination, we compared phospho-Spt5 levels in this region in control and Xrn2 D235A expressing cells where termination is inhibited. Although total Spt5 accumulates to high levels in 3’ flanking regions in Xrn2 D235A cells, it is not detectably phosphorylated (Fig. 4A, B, S3B). This experiment therefore strongly suggests that loss of Spt5 phosphorylation downstream of poly(A) sites is not dependent on termination.

To determine whether the switch from high to low Spt5 phosphorylation at 3’ ends requires poly(A) site recognition, we again made use of NS1A induction to inhibit CPSF, and monitored phospho-Spt5 and total Spt5 by ChIP. Notably, disruption of CPA by NS1A also impaired the apparent Spt5 dephosphorylation that occurs downstream of genes (Fig. 4G, H, S3C-E). In summary, the transcriptional slowdown in termination zones (Fig. 1 and 2) correlates with reduced phosphorylation of the Spt5 associated with ECs in this region, and this switch is dependent on poly(A) site recognition, but not termination.

PNUTS binding to PP1 is required for Spt5 de-phosphorylation at 3’ ends

The loss of Spt5 phosphorylation at 3’ ends of genes could be caused by exchange of phospho-Spt5 for a hypo-phosphorylated isoform or by phosphatase activity. Because the protein phosphatase PP1 and its binding partner PNUTS co-purify with 3’ processing complexes (Shi et al., 2009; Vanoosthuyse et al., 2014), we investigated a possible role of PNUTS-PP1 in loss of Spt5 phosphorylation at 3’ ends. To disrupt PNUTS-PP1 we induced expression of the PNUTS W401A protein, which is mutated in the PP1 interaction motif (Kim et al., 2003), and as a control, wild-type PNUTS was also over-expressed (Fig. S4A). The W401A dominant-negative mutant has the advantage that it distinguishes the phosphatase co-factor function of PNUTS from its other functions including RNA binding. We monitored Spt5 phosphorylation by normalizing phospho-Spt5 ChIP signals to total Spt5 after inducing expression of PNUTS W401A or the wild-type control. PNUTS W401A expression markedly increased Spt5 phosphorylation within gene bodies and 3’ flanking regions (Fig. 5A-D, S4F, G). Importantly, in PNUTS W401A cells, the profile of normalized phospho-Spt5 levels around poly(A) sites differed substantially from the wild type control. Instead of a precipitous fall immediately downstream of the poly(A) site, a smaller and more gradual loss of Spt5 phosphorylation occurred in the PNUTS mutant cells (Fig. 5C, S4F). Elevated Spt5 phosphorylation in PNUTS W401A expressing was not associated with a marked change in the amount of total Spt5 relative to pol II within gene bodies or 3’ flanking regions (Fig. 5A, B, S4D, E). Moreover, the effect of PNUTS W401A on Spt5 phosphorylation is not a secondary effect of cell cycle arrest as shown by flow cytometry of cells before and after doxycycline induction (Fig. S5A).

To investigate whether PNUTS-PP1 could directly dephosphorylate Spt5 at poly(A) sites, we used a substrate-trapping strategy (Wu et al., 2018). A fusion of PNUTS with catalytically dead PP1β (D63A) was expressed in HEK293 cells with an EGFP tag. The fusion protein was pulled down on GFP-trap beads and probed for Spt5. This experiment showed that Spt5 does indeed bind to EGFP-PP1(D63A)-PNUTS but not to EGFP alone (Fig. 5E) consistent with direct dephosphorylation of Spt5 by PNUTS-PP1. Also consistent with this conclusion, is the fact that in vivo bulk Spt5 becomes hyperphosphorylated in response to high concentrations of the phosphatase inhibitor okadaic acid as evidenced by conversion to a slower migrating species (Fig. S4C).

To ask whether PNUTS-PP1 is present at 3’ ends where Spt5 phosphorylation is lost from EC’s, we mapped wild type and mutant PNUTS proteins globally by ChIP. Epitope tagged wild-type PNUTS strongly localized in discrete peaks just downstream of poly(A) sites, close to where Spt5 dephosphorylation occurs (Fig. 5F, S5B-D). A substantial amount of PNUTS was also present at 5’ ends of many genes. Notably, the peak of PNUTS density downstream of poly(A) sites is distinct from that of pol II CTD Ser2 phosphorylation which is broader and shifted further downstream (Fig. S5D). By contrast, the PNUTS W401A protein was more diffusely localized than the wild type with a stronger preference for 5’ ends (Fig. 5F, S5B, C). The delocalization of PNUTS from 3’ ends as a result of disrupting PP1 binding, suggests that the phosphatase activity is also probably no longer concentrated at 3’ ends. In summary, the PNUTS protein closely co-localizes with the site of dephosphorylation, and its mis-localization in the W401A mutant is associated with poor Spt5 dephosphorylation (Fig. 5F). Together these experiments suggest a direct role of PNUTS-PP1 in limiting Spt5 phosphorylation within gene bodies and in the sharp transition to a Spt5 hypo-phosphorylated state downstream of poly(A) sites.

PNUTS binding to PP1 is globally required for transcription termination

Previously PNUTS was found to be necessary for efficient transcriptional termination at a subset of genes using a knockdown approach (Austenaa et al., 2015), but it remained unclear if this function involved regulation of PP1 activity. We therefore examined whether specific disruption of the PNUTS-PP1 interaction by the W401A mutant has a termination phenotype. Consistent with a vital role for PNUTS-regulated PP1 activity in termination, we observed that the W401A mutant suppressed termination globally as demonstrated by elevated levels of total pol II extending several kilobases downstream of genes (Fig. 5A, B, G). PNUTS W401A not only inhibited termination and Spt5 dephosphorylation at genes encoding poly(A)+ mRNAs, but also at replication-dependent histone genes that make poly(A)− mRNAs (Fig. S6A-D). A role for PNUTS in termination at histone genes is also consistent with its strong localization at the 3’ ends of these genes (Fig. S6E).

The transcription termination defect caused by PNUTS-PP1 disruption is not associated with defective CTD Ser2 hyperphosphorylation that normally occurs when pol II reaches the poly(A) site (Fig. S4B) (Buratowski, 2009). This result suggests that PNUTS W401A does not interfere with pre-mRNA 3’ end formation coupled to Ser2 phosphorylation (Davidson et al., 2014). To test this idea, we measured the effect of PNUTS W401A on cleavage at poly(A) sites. Bromouridine pulse-labeled nascent RNA-seq libraries were prepared before and after PNUTS W401A induction and uncleaved reads that overlapped poly(A) cleavage sites were quantified relative to those immediately upstream (see Star Methods). This analysis showed no detectable inhibition of poly(A) site cleavage by PNUTS W401A (Fig. 5H). In fact, a small increase in cleavage efficiency was observed, whose significance is unclear. We also asked whether PNUTS W401A affects recruitment of the Xrn2 terminator exonuclease. Anti-Xrn2 ChIP-seq showed that the exonuclease associated with ECs that failed to terminate in the presence of PNUTS W401A (Fig. S6F, G). Nor is disruption of PNUTS-PP1 likely to inhibit the Xrn2 exonuclease activity since its phosphorylated isoform is actually more active (Sanso et al., 2016). In summary, these results argue that the termination defect caused by PNUTS W401A is not an indirect consequence of impaired 3’ end processing or Xrn2 activity. Instead, disruption of the PNUTS-PP1 complex effectively uncouples termination from poly(A) site cleavage and torpedo loading.

PNUTS-PP1 functions as a brake on pol II elongation throughout the genome

How does disruption of PNUTS-PP1 inhibit termination independently of cleavage/polyadenylation? We hypothesized that PP1 phosphatase activity is required to decelerate the EC downstream of poly(A) sites, which would promote termination by allowing the Xrn2 torpedo to easily catch up with the polymerase and dislodge it from the DNA template. To test this model, we measured elongation rates by the DRB-release method in cells overexpressing PNUTS W401A or wild type PNUTS as a control. Strikingly, the PNUTS W401A mutant caused a genome-wide acceleration of pol II transcription elongation that was evident both in gene bodies and 3’ flanking sequences (Fig. 6, S7). Median elongation rates within gene bodies increased from 2.2 kb/min in PNUTS wild type cells to 3.6 kb/min in the W401A mutant (Fig. 6D). In 3’ flanking regions PNUTS W401A expression increased pol II speeds more than three-fold, from a median rate of 0.7 kb/min to 2.4 kb/min (Fig. 6D). Note that in PNUTS W401A cells, after the pol II wave front reaches the poly(A) site (5 min. time point Fig. 6D), it continues farther downstream at later time points, whereas in control cells it persists close to the poly(A) in a stable peak of pol II density (Fig. 6D, S7, red arrows). Suppression of the 3’ peak of pol II by PNUTS W401A is also evident in snapshots of pol II density (Fig. 5A, B, G). PNUTS W401A also suppressed pol II accumulation at apparent pause sites within genes (Fig. 6A, S7A, black arrows) suggesting that it may accelerate overall elongation rates by suppressing pausing. Importantly, the strong acceleration of transcription in 3’ flanking regions elicited by PNUTS W401A suggests that this phosphatase complex is normally responsible for triggering the transition to slow elongation at the poly(A) site (Fig. 1, 2), and that specific DNA sequences within termination zones are not generally sufficient to slow down transcription. The ability to experimentally accelerate transcription with the PNUTS W401A dominant mutant provides an opportunity to test the model that slow elongation will cause pol II to pile up, creating regions of high occupancy (Ehrensberger et al., 2013). Consistent with this model, we found that reversing pol II deceleration at 3’ ends by expressing PNUTS W401A strongly depressed the peak of pol II density that occurs downstream of most poly(A) sites (Fig. 6B-D, S7B). PNUTS W401A did not strongly affect apparent promoter-proximal pol II pausing as measured by pol II density at the TSS versus the gene body (Fig. S5E, F). Note however, that if release from the 5’ pause were stimulated by Spt5 hyperphosphorylation under PNUTS W401A conditions, it may not be detected as a net increase in pol II density in the gene body because those polymerases travel at an accelerated rate (Ehrensberger et al., 2013). In summary these experiments show that the PNUTS-PP1 phosphatase exerts a global brake that limits pol II speed within genes; and downstream of genes, it imposes a strong deceleration that is necessary for efficient termination.

Control of elongation speed by PNUTS-PP1 and Spt5 phosphorylation

We investigated the regulation of RNA pol II speed along human transcription units. Using two experimental approaches, we found that pol II sharply decelerates downstream of poly(A) sites. Median elongation rates decreased from 2-3 kb/min within gene bodies to only 0.1-0.9 kb/min downstream of poly(A) sites (Fig. 1, 2). This slow down accounts for much of the pile up of pol II in 3’ flanking regions, which is a ubiquitous characteristic feature of metazoan genes (Anamika et al., 2012; Core et al., 2008; Davidson et al., 2014; Glover-Cutter et al., 2008; Gromak et al., 2006; Lian et al., 2008) because when transcription was experimentally accelerated, the pile up was suppressed (Fig. 6, S7B). The 3’ pol II pile up was also suppressed by blocking poly(A) site recognition, either by the influenza NS1A protein or by mutating the AATAAA consensus sequence (Fig. 3). These results therefore suggest that the association of CPA factors with the nascent transcript, or RNA cleavage itself, is required to switch the EC into a slowly elongating conformation. Such an allosteric switch that changes the elongation properties of pol II complexes was predicted over thirty years ago as a mechanism for poly(A) site dependent termination (Logan et al., 1987).

Three lines of evidence suggest that the switch to slow transcription is mediated at least in part by a poly(A) site-dependent dephosphorylation of the universal elongation factor Spt5 by PP1 phosphatase targeted by its binding partner PNUTS: First, changes in Spt5 phosphorylation at 5’ and 3’ positions coincide with changes in elongation rate. Thus Spt5 phosphorylation in EC’s increased after pol II release from the promoter-proximal pause site (Fig. 4F, 5D, S4G) when elongation accelerates (Danko et al., 2013; Jonkers et al., 2014), and then decreased within 100 bases downstream of the poly(A) site when elongation decelerates (Fig. 4C-E,G,H, 5C, S4F). Second, loss of phospho-Spt5 at 3’ ends requires poly(A) site recognition (Fig. 4G). Third, disruption of the PNUTS-PP1 complex by the PNUTS-W401A mutation increases Spt5 phosphorylation and globally accelerates transcription within genes and 3’ flanking regions (Fig. 6, S7). Thus PNUTS-PP1 negatively regulates transcription elongation rate at early and late stages of the transcription cycle. It will be interesting in future to examine how global acceleration of transcription by PNUTS-PP1 disruption affects co-transcriptional mRNA processing and chromatin modification. Our results do not exclude the possibility that PNUTS-PP1 regulates Spt5 phosphorylation indirectly, however its in vivo binding to Spt5 (Fig. 5E), the concentration of PNUTS polypeptide in a discrete peak immediately downstream of poly(A) sites, and the re-distribution of the PNUTS-W401A protein away from 3’ ends (Fig. 5F, S5B-D, S6E) all suggest a direct role at poly(A) sites. Of course, dephosphorylation of other potential PNUTS-PP1 substrates including the pol II CTD (Ciurciu et al., 2013; Lee et al., 2010) might also modulate elongation, but no CTD dephosphorylation has yet been identified specifically in the vicinity of poly(A) sites.

Control of elongation rate and the mechanism of transcription termination

The deceleration by at least 2-3 fold that occurs within ~100 bp downstream of poly(A) sites (Fig. 1, 2) is predicted to facilitate termination by converting the pol II EC into an easy target or “sitting duck” that can be quickly tracked down by the Xrn2 torpedo. In other words, pol II deceleration biases the kinetic competition between pol II and the Xrn2 torpedo that pursues it (Fong et al., 2015). The “sitting duck torpedo” mechanism of poly(A) site dependent termination comprises essential elements of both the torpedo and allosteric models that have been debated for over 30 years (Libri, 2015) (Fig. 7). We propose that termination requires both transcript cleavage that provides access to the Xrn2 exonuclease “torpedo” and poly(A) site-dependent deceleration of transcription that converts the polymerase into a viable target for the “torpedo” to track down and dislodge from the DNA template.

This hybrid model can account for the finding that poly(A) site cleavage is not always sufficient to induce termination (Osheim et al., 1999). It also explains the puzzling observation that there are distinct outcomes for pol II when termination is inhibited in different ways (Eaton et al., 2018). When the Xrn2 exonuclease is inactivated, polymerases still undergo poly(A) site-dependent deceleration so they do not travel far into the 3’ flanking region (Figs. 1, 2). In contrast, when 3’ processing is prevented by CPSF73 depletion (Eaton et al., 2018), NS1A expression, or elimination of the AAUAAA element (Fig. 3), then rapid elongation persists, the pol II pile up is circumvented, and ECs travel much further into the 3’ flanking region. Poly(A) site dependent termination can thus be disrupted in two distinct ways: 1. By impeding torpedo function either by restricting Xrn2 activity or its access to the transcript by preventing poly(A) site cleavage, and 2. By preventing pol II deceleration, for example through inhibition of Spt5 dephosphorylation at 3’ ends, so the torpedo cannot catch up with its target. It is possible that the strong termination defect caused by mutation of Tyr residues in the pol II CTD heptad repeats works by the latter mechanism (Shah et al., 2018).

Poly(A) site-dependent deceleration of pol II has potential implications for co-transcriptional alternative poly(A) site selection where upstream and downstream processing sites compete with one another (Mayr, 2016; Tian and Manley, 2017). If recognition of the poly(A) site converts pol II into its slow conformation, then this would extend the window of opportunity for completion of 3’ end formation at that site before competing downstream poly(A) sites are transcribed. One mystery that remains however is whether pol II deceleration requires cleavage of the nascent RNA or just assembly of a 3’ processing complex on the poly(A) site.

At poly(A) sites we propose that PP1 activity is up-regulated to dephosphorylate Spt5. PNUTS-PP1 forms part of the CPA apparatus (Shi et al., 2009; Vanoosthuyse et al., 2014) and its phosphatase activity could be activated or de-repressed by conformational changes associated with poly(A) site recognition or RNA cleavage. In fission yeast, Dis2 dephosphorylates Spt5 at the ends of genes but this transition was unaffected by mutation of the cleavage polyadenylation factor Pfs2, so its relationship to poly(A) site dependent termination is unclear (Kecman et al., 2018; Parua et al., 2018). This apparent discrepancy between human and fission yeast systems could reflect a mechanistic difference, or alternatively the pfs2 mutant may not block assembly of CPA complexes sufficiently to affect Spt5 dephosphorylation. Another difference between species is that in S. pombe Spt5 dephosphorylation is thought to cause its displacement from pol II by the termination factor Seb1 (Kecman et al., 2018) whereas in human cells dephosphorylated Spt5 persists in association with pol II throughout the termination zone (Fig. 4A-E, S3B). Furthermore, the association of Spt5 with slow pol II elongation complexes downstream of poly(A) sites shows that binding of this factor is not always sufficient to promote rapid elongation (Baluapuri et al., 2019).

Similar roles for Spt5 and NusG in control of transcriptional speed and termination

How might the phosphorylation status of Spt5 flip a switch between slow and fast modes of elongation? Spt5 dephosphorylation could alter the conformation of the EC directly or alternatively it could work indirectly through recruitment or release of an associated factor. Phosphorylation of Spt5 could relieve a brake exerted by binding of the unphosphorylated protein to pol II. One possibility is that dephosphorylation enhances Spt5 binding to the nascent transcript, which might in turn slow down transcription. In this context it is interesting to note that a sharp increase in RNA binding was detected by Spt5 PAR-CLIP immediately downstream of poly(A) sites in budding yeast (Baejen et al., 2017). CTR1 phosphorylation might also influence elongation rate by regulating interaction with other elongation/termination factors like the Rtf1 subunit of the Paf1 complex (Wier et al., 2013; Zhou et al., 2009) that synergizes with TFIIS (Kim et al., 2010) to stimulate elongation by permitting backtracked complexes to resume transcription. Pol II backtracking is a major limiting factor for transcription elongation in human cells, and the frequency of backtracking is elevated downstream of poly(A) sites, but the basis for this phenomenon is unknown (Sheridan et al 2019). It seems likely that the switch to slow elongation at poly(A) sites might involve an increased propensity to backtrack. Interestingly the NusG homologue of Spt5 controls backtracking of E. coli RNA polymerase by altering the conformation of the transcription bubble (Turtola and Belogurov, 2016). It is also possible that Spt5 dephosphorylation alters the pol II active site conformation (Vos et al., 2018b) and induces slow elongation independently of backtracking. It will be of interest in future to investigate how the PNUTS-PP1 phosphatase affects pol II elongation at high resolution. The poly(A) site activated brake on transcription mediated through Spt5 likely reflects an ancient dual role of this protein family in coupling control of elongation rate with termination. In bacteria NusG family members regulate termination by the Rho helicase (Artsimovitch and Knauer, 2019; Ray-Soni et al., 2016), which chases RNA polymerase in an analogous fashion to the Xrn2 exonuclease. We speculate that hypo- and hyper-phosphorylated isoforms of Spt5 in eukaryotes enhance or suppress termination like NusG in bacteria by biasing the kinetic competition between polymerases and the factors that pursue them, Xrn2/Rat1 and Rho respectively. In addition to control of termination, it is also possible that PNUTS-PP1 regulation of elongation rate could occur within gene bodies in response to intrinsic or environmental stimuli.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, David Bentley (david.bentley@cuanschutz.edu)

EXPERIMENTAL MODEL AND SUBJECT DETAILS

METHOD DETAILS

Quantification and Statistical Analysis

Statistical details are provided in figure legends. The number of replicates and the correlations between them are provided in Table S1. Significance was defined by p-values determined by using the Welch t-test. When appropriate, p-values were corrected for multiple testing using the Bonferroni-Holm method.

Data and Code Availability

Sequencing Data are available at GSE134198.