mRNA decapping factors and the exonuclease Xrn2 function in widespread premature termination of RNA polymerase II transcription

Termination and decapping factors localize at 5′ ends near paused pol II

We investigated the functional significance of the co-purification of decapping factors and TTF2 with Xrn2. The association of decapping factors with a nuclear enzyme Xrn2 was surprising as decapping is thought to be predominantly cytoplasmic (Franks and Lykke-Andersen, 2008; Song et al., 2010). However, we found that decapping activity is readily detectable in Hela nuclear extract and can be specifically immunoprecipitated by anti-Dcp1a antibody (Fig. S1A). We investigated whether decapping factors might act in the nucleus at sites of transcription by ChIP-Seq analysis of Dcp1a and pol II in Hela cells. Because Dcp1a interacts with TTF2 (Leonard et al., 2003) and Xrn2 (Fig. 1), we further tested whether these three factors co-localize on genes by ChIP-seq with antibodies whose specificity was confirmed by shRNA knockdown of the respective proteins (Fig. S1B, (Jiang et al., 2004)). The Dcp1a ChIP-seq signals were sufficient to identify 745 genes with peaks of enrichment within 500 bases of a TSS (FDR < 0.05, Table S2) detected by the HOMER peak finder (Heinz et al., 2010). Note that while the peak finding analysis of our ChIP-seq results reveals significant enrichment of Dcp1a, Xrn2 and TTF2 near TSSs, it does not provide a complete list of genes where these factors are found at the 5′ end. Furthermore Xrn2 and TTF2 co-localized with Dcp1a near these start sites to a significant extent (Figs. 1C, D, Tables S3, S4). Interestingly peaks of Xrn2 and TTF2 accumulation (FDR <0.05) were more frequent within 500 bases of a TSS than they were in the region 0–3kb downstream of a poly(A) site (Table S2). Importantly, the ChIP signals for Dcp1a, TTF2 and Xrn2 overlap extensively with one another and with pol II at several thousand genes in the region between −1kb and +2kb from the TSS (Figs. 1D, S2E, see also Tables S3, S4). We confirmed the localization of Xrn2 at 5′ ends by ChIP in two human breast cell lines in addition to Hela (Fig. S2A). In addition when transcription elongation was inhibited with DRB, the distribution of Xrn2 on the GAPDH gene shifted toward the 5′ end of the gene in parallel with pol II (Fig. S2B, C). The localization of TTF2 at 5′ ends of genes was further confirmed by ChIP-seq using an independent commercial polyclonal antibody (Fig. S2D). We conclude that these termination and decapping factors commonly localize near the 5′ ends of genes.

Depletion of Xrn2 and TTF2 re-distributes pol II away from the TSS toward promoter-distal positions

Co-localization of a decapping factor with Xrn2 at 5′ ends suggests a possible parallel with cytoplasmic mRNA degradation by decapping and 5′-3′ exonucleolytic degradation by Xrn1 (Coller and Parker, 2004). According to this model, decapping of nascent transcripts coupled to Xrn2-mediated degradation might cause premature termination of transcription by the “torpedo” mechanism. If decapping factors, Xrn2 and TTF2 terminate a population of promoter-proximal polymerases, then their depletion would enhance readthrough transcription downstream and/or upstream of promoters. In this event relative pol II density in promoter-distal regions is expected to increase relative to promoter-proximal regions. To test this prediction, we first knocked down Xrn2 and TTF2 individually and together in stable HEK293 cell lines by infection with lentiviral shRNA expression vectors (Fig. S1B, Table S5). We then measured pol II occupancy along genes by ChIP-Seq in the knockdown lines relative to two controls, the uninfected parent and a line infected with a scrambled shRNA lentivirus. Knockdown of Xrn2 and/or TTF2 had only small effects on the overall enrichment of pol II within genes compared to parent or scrambled shRNA controls, as determined by anti-pol II ChIP-seq reads per kilobase per million in the region from −30 to the poly(A) site (Fig. 2B, S3C). The depletion of Xrn2 and TTF2 achieved in our cell lines therefore does not cause widespread inhibition of transcription initiation. Pol II density profiles normalized to total read-counts (Fig. 2A) revealed that, as predicted by the model, on genes where pol II normally accumulates in the promoter-proximal region, knockdown of Xrn2+TTF2 markedly increased pol II density at promoter-distal positions and decreased it in promoter-proximal positions. Knockdown of Xrn2 or TTF2 individually, caused a similar redistribution of pol II from promoter-proximal to promoter-distal positions, but the effects were smaller than in the double knockdown (Fig. S3A). The complementary changes in pol II density at distal and proximal sites are consistent with a redistribution of polymerase away from the start site and into the body of the gene. In contrast only modest effects on pol II 5′-3′ distribution were evident on genes with lower levels of promoter-proximal pol II accumulation including HIST1H4C, SFN, and ARHGDIA and the non-coding RNA gene HOTAIR (Fig. S4A–D).

Open in a separate window

Figure 2

Knockdown of Xrn2 and TTF2 increases relative pol II occupancy upstream and downstream of start sites

(A) Pol II ChIP-Seq normalized to total read counts on indicated genes in HEK293 cells stably expressing shScramble (scr), and shXrn2+shTTF2 shRNAs. Note reduced pol II density near the TSS and increased density within the gene and 3′ flank with knockdown of Xrn2+TTF2 (see Fig. S2A for individual knockdowns). Blue arrow indicates divergent transcription at RBM39 (see Fig. S5D) that is slightly elevated by Xrn2+TTF2 knockdown. Images were made using the IGB6.4 Browser (Nicol et al., 2009). (B) Pol II densities expressed as log2 RPKM (reads per kilobase per million) in the region from −30 to the poly(A) site (see diagram) in the shScrambled (scr) control relative to uninfected parent and double knockdown lines (see also Fig. S2C). Each plot represents a set of 5507 genes >2kb long and >2kb from neighbors enriched for pol II (FDR<.05) within 500 bases of a TSS that are shared in common between the uninfected parent, scrambled control and knockdown datasets. We noted some decrease in RPKM in knockdown lines among more highly expressed genes but the number of genes that significantly diverge from the scrambled control (p<.01, two-tailed T test) is a small fraction of the total.

On the RBM39 gene where divergent antisense transcription occurs (Fig. S5D) (Core et al., 2008; Preker et al., 2008), depletion of Xrn2+TTF2 modestly increases pol II density upstream of the TSS as well as downstream (Fig. 2A). This observation suggests that Xrn2 and TTF2 may normally limit the amount of divergent antisense transcription. Similar enhancement of antisense transcription in the Xrn2+TTF2 knockdown cells occurred at the ARHGDIA, ID1 and FRAT2 genes (Fig. S4D–F). These observations on individual genes were confirmed in an average of many genes (see Figs 3A, B).

Open in a separate window

Figure 3

Knockdown of Xrn2 and TTF2 shifts the distribution of pol II away from the TSS

(A) ChIP-Seq relative frequency profiles (reads per 50bp bin divided by total reads in all bins, see Experimental Procedures) in shScramble (scr) control and shXrn2+shTTF2 expressing HEK293 cells across 5507 genes used in Fig. 2B (B) Log2 ratio of pol II ChIP-Seq read frequencies in uninfected parent and knockdown cell lines relative to the shScramble (scr) control. Note the additive effect with Xrn2+TTF2 knockdown. (C) Definition of Escape Index (EI) (D) Escape index (EI) for the 5507 genes analyzed in Fig. 2B (Table S2). The numbers of genes that differ from the scr control (FDR < 0.01, one-sample T-test) are given. Note widespread elevation of EI in the Xrn2:TTF2 knockdown line. Best fit lines (orange) were generated by loess fitting (Local Polynomial Regression Fitting). Black dots correspond to genes with significant (FDR <.05) 5′ peaks of Dcp1a ChIP signal in Hela (Table S2).

Xrn2 and TTF2 shRNAs did not cause a major inhibition of transcription termination downstream of poly(A) sites (Fig S5E) at the level of knockdown achieved in our cells consistent with a previous report (Banerjee et al., 2009). At some genes like ACTB however termination was delayed (Fig. 2A). In summary the results in Figure 2A show that depletion of two termination factors that localize at 5′ ends of genes caused a shift in the distribution of pol II from promoter-proximal to promoter-distal positions both upstream and downstream of the TSS.

To determine how generally Xrn2/TTF2 depletion affects transcription within human genes, we assessed pol II distribution on a cohort of 5507 genes in each knockdown cell line (Table S2). The genes analyzed were selected on the basis that they are >2kb long and >2kb from neighboring genes and are significantly enriched (FDR<0.05) for pol II in the region within 500 bases of a TSS. First we plotted the relative frequency of pol II ChIP-seq reads on these genes in the region from −1kb to +2kb from the TSS for the Xrn2+TTF2 double knockdown and scrambled shRNA control lines (Fig. 3A). This analysis shows that the relative increase in pol II read frequency at positions both upstream and downstream of the TSS in the double knockdown line is detectable even when averaged over more than 5000 genes. To quantify the effects of Xrn2 and TTF2 depletion on a broad scale, we plotted the ratios of the relative frequencies of pol II ChIP-seq reads (see Experimental Procedures) in each knockdown cell line compared to the scrambled shRNA control (Fig. 3B). As a control we compared the uninfected parent line to the scrambled shRNA (grey line Fig. 3B) and as expected there was little difference. In contrast, each of the knockdown lines demonstrated a specific increase in the relative frequency of pol II at promoter-distal positions compared to the scrambled control with the greatest effect in the Xrn2:TTF2 double knockdown (Fig. 3B). The apparent additive effect of depleting TTF2 and Xrn2 is significant because the only known function in common between these proteins is in transcription termination. Notably depletion of termination factors increased relative pol II frequency at positions both downstream and upstream of the TSS and these effects are widespread as evidenced by the fact that they are detectable in plots averaged over the group of 5507 genes.

To investigate further how termination factors influence pol II distribution across genes, we calculated the escape index (EI), defined as log2 promoter-distal (+301 to the polyA site) pol II density: promoter-proximal (−100 to +300) pol II density (Fig. 3C) in control and knockdown HEK293 cell lines for 5507 genes. This analysis, revealed significant (FDR< 0.01) increases in EI for many genes in the Xrn2+TTF2 double knockdown line, but not the uninfected parent cells, relative to the scrambled shRNA control (Fig. 3D). Importantly the greatest effects of Xrn2/TTF2 depletion on EI were evident for those genes with low EI values that have the greatest accumulation of promoter-proximal pol II density. In contrast genes in the minority class of human genes with high EI values and low levels of promoter-proximal pol II accumulation, were less affected by depletion of these termination factors. This point is particularly clear for the Xrn2:TTF2 double knockdown (Fig. 3D, red) and is further supported by inspection of individual genes with low levels of promoter-proximal pausing and high EI values (Fig. S4A–D). In summary, the results in Figure 2 and ​and33 show that at many genes the termination factors Xrn2 and TTF2 limit pol II elongation away from the TSS in both directions, consistent with widespread premature termination in promoter-proximal regions.

Decapping factor knockdown re-distributes pol II away from promoter-proximal positions

If Xrn2 and TTF2 cooperate to facilitate termination at promoter-proximal regions by a “torpedo” mechanism, then decapping or RNA cleavage would be required to provide a suitable 5′ phosphorylated substrate for the exonuclease. To test whether decapping factors affect how pol II is distributed between promoter-proximal and distal regions, we stably knocked down Edc3, Dcp1a or Dcp2 in HEK293 cells (Fig. S1C) and analyzed pol II distributions by ChIP-Seq. Similar to the termination factors, depletion of Dcp2, Edc3, or Dcp1a did not cause major changes in overall levels of pol II enrichment within genes when compared to the scrambled shRNA control (Fig. 4B, S3D). Depletion of all these decapping factors, especially Dcp2, specifically increased pol II densities at promoter-distal locations and decreased it at the TSS (Figs. 4A, S3B). As we observed for Xrn2/TTF2, depletion of decapping factors enhanced relative pol II accumulation at the 3′ pause that precedes termination at some genes like ACTB probably because more polymerases reach the 3′ end (Fig. 4A). Knockdown of decapping factors increased pol II relative frequency at promoter-distal positions both upstream and downstream of the TSS (Fig. 5A, B). In addition, plots of escape index (EI, Fig. 3C) for the group of 5507 genes used in Figure 3 revealed that the promoter-proximal to -distal shift in pol II distribution caused by depleting decapping factors is widespread (Fig. 5C). In summary, these results reveal a surprising effect of depleting decapping factors on pol II transcription that strongly resembles the effect of depleting the termination factors Xrn2 and TTF2. We interpret these results to suggest that decapping, coupled to polymerase displacement by the Xrn2 “torpedo” and the ATPase TTF2, facilitates premature termination of bidirectional transcription from human promoters.

Open in a separate window

Figure 4

Knockdown of decapping factors increases relative pol II occupancy upstream and downstream of start sites

(A) Pol II ChIP-Seq normalized to total read counts in HEK293 cells stably expressing shScramble (scr) and shDcp2 shRNote reduced pol II density near the TSS and increased density within the gene and 3′ flank with knockdown of Dcp2 (see Fig. S2B for Edc3 and Dcp1a knockdowns). Blue arrow indicates divergent transcription at RBM39 (see Fig. S5D) that is slightly elevated by Dcp2 knockdown. (B) Pol II densities expressed as log2 RPKM (reads per kilobase per million) in the region from −30 to the poly(A) site in the scr control relative to uninfected parent and shDcp2 as in Fig. 2B (see also Fig. S2C). Note knockdown of Dcp2 does not have a major effect on total pol II occupancy within genes.

Open in a separate window

Figure 5

Knockdown of decapping factors and Spt5 cause similar re-positioning of pol II away from the TSS

(A) ChIP-Seq relative frequency profiles (5507 genes) in shScramble (scr) and shDcp2 expressing HEK293 cells as in Fig. 3A. (B) Log2 ratio of pol II ChIP-Seq read frequencies in uninfected parent and knockdown cell lines relative to the scr control as in Fig. 3B. (C) Escape index (EI, see Fig. 3C) in knockdown lines and uninfected parent compared to scr control as in Fig. 3D. The numbers of genes that differ from the scr control (FDR < 0.01, one-sample T-test) are given. Note widespread elevation of EI particularly in the Dcp2 knockdown line. Black dots correspond to genes with significant 5′ peaks of Dcp1a ChIP signal as in Fig. 3D.

(D) ChIP-Seq relative frequency profiles as in (A) in mouse ES cells expressing shScramble (scr) (6813 genes) shNelfA (6467 genes) and shSpt5 (5604 genes) shRNAs from (Rahl et al., 2010). (E) Log2 ratios of pol II ChIP-Seq read frequencies in shNelfA and shSpt5 expressing ES cells relative to the scr control. (F) Escape index (EI) in shNelfA and shSpt5 ES cells relative to the scr control. FDR was not calculated because only one control, the scrambled shRNA, was analyzed. Note elevation of EI (F) and pol II relative frequency in the gene body (D) in Spt5 knockdown cells is comparable to decapping factor knockdown cells (A–C).

Comparing the roles of termination, decapping and pausing factors in pol II localization

The current model for control of pol II transcriptional elongation on a large fraction of metazoan genes is by regulated release from a promoter-proximal pause that is established by NELF and DSIF (Spt4/5) (Chiba et al., 2010; Core and Lis, 2008; Levine, 2011; Nechaev and Adelman, 2011; Peterlin and Price, 2006). Previously the effects of depleting NelfA and Spt5 on transcription were defined by pol II ChIP-Seq in mouse embryonic stem (ES) cells (Rahl et al., 2010) and we compared these results with the effects of depleting decapping and termination factors. In agreement with Rahl et al (2010), pol II relative frequencies and EI on over 5000 mouse genes showed that knockdown of Spt5 increased pol II density at downstream positions relative to the TSS (Fig. 5D, E) just as we saw for knockdown of termination and decapping factors (Figs. 3A,B, 5A,B). The magnitude of the increase in relative pol II density within genes caused by Spt5 knockdown was similar to that caused by Edc3, Xrn2, or TTF2 knockdown and less than Dcp1a, Dcp2 or Xrn2:TTF2 double knockdown. In contrast with the latter factors, Spt5 knockdown appeared to have little effect on divergent transcription as judged by pol II densities upstream of the TSS while the NelfA knockdown had a much more modest effect on average pol II distribution among these genes (Fig. 5E) as previously reported (Rahl et al., 2010). Note that Spt5 knockdown resulted in a step like increase in pol II density immediately downstream of start sites whereas termination and decapping factor knockdown causes a gradual increase in relative pol II density over a 2kb region downstream of the start site. This effect is consistent with premature termination by the “torpedo” occurring over hundreds of bases, similar to termination downstream of genes.

Antibodies

Rabbit anti-pan pol II CTD was described previously (Schroeder et al., 2000). Anti-Xrn2 was raised in rabbits against GST-Xrn2 (a.a. 402–537) and affinity purified. Affinity purified rabbit anti-GFP (Fong et al., 2009), rabbit anti-Dcp1a, and anti-Edc3 have been described (Fenger-Gron et al., 2005; Lykke-Andersen and Wagner, 2005). Anti-Dcp2 and Anti-Aly (11G5, THOC4) were gifts of M. Kiledjian and G. Dreyfuss. Rabbit anti-TTF2 was from Proteintech and affinity purified sheep anti-TTF2 was a gift of D. Price.

Xrn2 immunopurification

Rabbit anti-Xrn2 and anti-GFP control antibodies were affinity purified and coupled to Amino-Link resin (Pierce) at 1 mg/ml and equal amounts were used for precipitation from HeLa Nuclear extract in the presence of RNAseA (20μg/ml) in buffer D (see Extended Experimental Procedures). IP’ed material was fractionated by 4–15% SDS-PAGE and each lane was cut into 14 bands. Proteins were reduced, alkylated and trypsin digested in the gel, extracted and analyzed by LCMS/MS on two analytical platforms: ThermoFisher LTQ XL and LTQ-FT Ultra. Results for the proteins most enriched in the Xrn2 IP relative to the GFP control by number of assigned spectra are listed in Table S1. Total assigned spectra were 22346 and 29540 for the anti-GFP, and 16776 and 23442 for the anti-Xrn2 LTQ and FT analyses respectively.

RNAi-mediated knock-down

pLKO.1-puro shRNA lentiviruses (Open Biosystems, Table S5) were used to infect HEK293-Flp-in T-REX-glob cells. For Xrn2:TTF2 double knockdown, a neomycin resistance marker was cloned in place of the puromycin resistance marker in pLKO.1-puro TRCN0000049900 targeting Xrn2 and knockdowns were verified by western blot (Fig. S1B).

ChIP-seq

ChIP and Illumina library preparation have been described (Kim et al., 2011). Libraries were sequenced on the Illumina Genome Analyzer IIx and Hi-Seq platforms. Single-end 34 base reads (after removing barcodes) were mapped to the hg18 UCSC human genome (Mar. 2006) with Bowtie version 0.12.5 (Langmead et al., 2009)(Table S6). We generated bed and wig profiles using 50bp bins and 200bp windows assuming a 180bp fragment size shifting effect. Results were viewed with the UCSC genome browser, integrated genome browser (IGB), or the R statistical software package.

This work was supported by NIH grant GM063873 to D.B. K.B. and K.G-C. were supported by Ruth L. Kirschstein NRSA awards F31GM095249 and F31GM72099. K.B. was supported by NIHT32-GM08730. S.K. was supported by the American Cancer Society and the Clark family (award PF-07-297-01-GMC). H.K. was supported by ARRA award 3R01GM063873-06S1, R.E.D. was supported by NIH grant AI49558; K.H was supported by NIH grant P30 CA046934-17 and NCRR grant S10RR023015. We thank U. Colorado Functional Genomics Core for shRNA constructs, M. Kiledjian (Rutgers U.), G. Dreyfuss (U. Penn) and D. Price (U. Iowa) for generous gifts of antibodies, T. Blumenthal, S. Johnson, T. Evans, J. Kim and R. Perales for helpful suggestions and J. Dover, D. Farrell, J. Castoe, B. Gao and K. Diener for Illumina sequencing.