Integrator Regulates Transcriptional Initiation and Pause Release Following Activation

SUMMARY

Determining the factors regulating the rate limiting steps in transcriptional control is of fundamental importance to understanding the mechanisms that govern eukaryotic transcription. While studies in unicellular organisms have pointed to initiation as the rate-limiting step in transcription, a large body of work in metazoans indicates that the transition to productive transcriptional elongation may also constitute a critical step. Here, we show that the RNA polymerase II (RNAPII)-associated multi-protein complex, Integrator, plays a critical role in both initiation and the release of paused RNAPII in immediate early genes (IEGs) following transcriptional activation by epidermal growth factor (EGF) in human cells. Integrator is recruited to the IEGs in a signal-dependent manner and is required to engage the super elongation complex (SEC) in pause release. We propose a role for Integrator as an RNAPII-associated factor modulating both initiation and pause release during transcriptional activation in metazoans.

INTRODUCTION

There is overwhelming evidence that transcriptional regulation of metazoan genes is controlled at two critical steps. The first involves the recruitment of sequence-specific DNA-binding transcription factors, the basal transcription machinery including RNA polymerase II (RNAPII) and general co-activator complexes to the promoter of responsive genes (Roeder, 2005; Sims et al., 2004). This phase culminates in the formation of competent transcriptional pre-initiation complexes leading to the initiation of transcription. However, accumulating evidence in metazoans using model systems such as Drosophila melanogaster and mammalian cells have indicated that in nearly 50% of genes, there is a second rate-limiting step about 20 to 60 nucleotides (in mammalian cells) down-stream of transcriptional start sites involving the transition of RNAPII to productive transcriptional elongation (Guenther et al., 2007; Kwak and Lis, 2013; Muse et al., 2007; Zeitlinger et al., 2007). At these genes, RNAPII experiences a barrier to productive elongation leading to what has been described as “paused” RNAPII. The precise molecular underpinning of this barrier to transcription has not been fully elucidated. However, accumulating evidence from studies on individual genes in vitro and in vivo, suggest a role for two transcription factors, NELF and DSIF, in the pausing mechanism (Lee et al., 2008; Wada et al., 1998; Yamaguchi et al., 1999). Moreover, it has been proposed that recruitment of the positive transcription elongation factor (P-TEFb), which is composed of the CDK9 kinase and Cyclin T, leads to the release of paused RNAPII into a productive elongation (Peterlin and Price, 2006), as a result of phosphorylation of NELF, DSIF and of the serine 2 residue of the C-terminal domain repeats of the largest subunit of RNAPII.

The detailed mechanism by which P-TEFb is recruited to transcriptionally active genes is not clear. There have been proposals suggesting that transcription factors such as Myc or RelA can directly recruit P-TEFb (Barboric et al., 2001; Eberhardy and Farnham, 2002; Rahl et al., 2010). Moreover, other transcriptional regulators, including BRD4 and the MED26 subunit of the Mediator complex, have been implicated in the recruitment of P-TEFb to specific target sites (Galbraith et al., 2013; Jang et al., 2005; Takahashi et al., 2011; Yang et al., 2005). Recent biochemical experiments have demonstrated the presence of an active P-TEFb in a larger multi-subunit complex termed Super Elongation Complex (SEC) containing other elongation factors such as ELL2 protein (Lin et al., 2010; Luo et al., 2012), suggesting the involvement of additional components of SEC in transcriptional pause release. However, aside from these examples, there has not been a unifying mechanism by which P-TEFb is recruited to genes following transcriptional activation leading to release of RNAPII from the paused state.

Among many interaction partners, two distinct multi-protein complexes, the Mediator and Integrator, are in a stable association with mammalian RNAPII through the C-terminal domain (CTD) of its largest subunit RPB1 (Baillat et al., 2005; Conaway and Conaway, 2011; Malik and Roeder, 2010). While Mediator is conserved throughout all eukaryotes, Integrator is only present in multi-cellular organisms. Moreover, while Mediator is thought to be present at mRNA genes, Integrator has only been implicated in snRNA biogenesis. Here, our genome-wide analysis has revealed an unexpected but critical role for the Integrator complex in the regulation of initiation as well as pause release at immediate early genes (IEGs), which are known for their regulation at the elongation step. We show that in a stimulus-dependent manner, Integrator is required for the recruitment of SEC-containing P-TEFb, leading to the release of paused RNAPII and resumption of productive elongation. We have extended these observations to demonstrate an analogous role for Integrator in heat shock gene activation in Drosophila revealing that its function in pause release is evolutionary conserved. Collectively, these results shed new light on our understanding of the basic mechanisms governing transcriptional regulation and also identify a new role for the Integrator complex in coordinating transcriptional elongation.

RESULTS

Integrator complex is recruited to paused genes

The Immediate Early Gene (IEG) and proto-oncogene c-Fos is the prototypical mammalian gene regulated through pause release (Plet et al., 1995). However, the scope of transcriptional regulation of IEGs through pause release has not been fully elucidated. The genes subject to pause release mechanism display a diagnostic signature with a predominant RNAPII peak proximal to their transcriptional start sites prior to transcriptional activation (Guenther et al., 2007; Muse et al., 2007; Zeitlinger et al., 2007). To identify additional IEGs regulated through RNAPII pause release following transcriptional activation, we used antibodies against the N-terminus of the largest subunit of RNAPII to perform chromatin immunoprecipitation followed by high throughput sequencing (ChIP-Seq). We used epidermal growth factor (EGF), a potent stimulating signal for IEGs to assess pause release in HeLa cells (Amit et al., 2007). EGF stimulation resulted in activation of 76 genes as measured by RNA sequencing (RNA-Seq) (Table S1). Nearly all IEGs responsive to EGF induction displayed a peak of proximal RNAPII reflective of paused RNAPII at their initiation sites prior to their activation (Figure 1A-C). Following EGF induction, these genes released their RNAPII into productive elongation as measured by analyzing the RNAPII traveling ratio (Figure 1D), a reliable measure of pause release comparing RNAPII occupancy on the promoter to that on the body of the gene (Rahl et al., 2010). Additionally, EGF stimulation resulted in increased recruitment of RNAPII as shown by RNAPII profile at 5′-end of EGF responsive genes (Figure 1A and B). Taken together, these results indicated that while EGF stimulation resulted in increased recruitment of RNAPII, a measure of enhanced initiation, EGF-responsive genes also displayed augmented transcriptional elongation as seen by RNAPII fold increase in the body of the genes (Figure 1C).

Figure 1. Integrator is recruited to immediate early genes following EGF induction.

(A) ChIP-seq tracks of RNAPII in HeLa cells before and after 20 minutes of EGF stimulation. Serum starvation (48h) causes accumulation of paused RNAPII at the TSS of immediate early genes such as CCNL1 and FOS. EGF stimulation releases RNAPII from the proximal promoter into the gene body. ChIP-seq tracks are visualized in a BigWig format and aligned to the hg19 assembly of the UCSC Genome Browser. (B) Average profile of RNAPII across 76 EGF-responsive genes, mean density is calculated as the average read number normalized to sequencing depth (total bin number: 240, from −1kb before the TSS to +3kb after the TES). (C) Fold recruitment of RNAPII after EGF stimulation. Fold increase is calculated as the ratio of the average read distribution before and after EGF (the average gene locus was divided in 35 bins from −500 upstream the TSS to 1kb downstream the TES). (D) Traveling ratio of RNAPII before and after EGF stimulation at 76 EGF-responsive genes. The ratio is calculated as log10 of read density at TSS/read density over the gene body. (E) qChIP analysis of Integrator recruitment at the TSS and 3′ end of JUN, FOS and NR4A1 genes. Three Integrator subunits (INTS1, INTS9 and INTS11) were examined during the time-course of EGF activation (0, 20, 40 and 60 minutes) in wild type HeLa cells. Averages of three experiments is shown.

(F) INTS11 is recruited to FOS, JUN, NR4A1 and CCNL1 in an EGF-dependent manner in wild type HeLa cells. ChIP-seq tracks of INTS11 before and 20 minutes following EGF stimulation show recruitment of INTS11 at the TSS and gene body of these four immediate early genes. Before EGF treatment cells were starved for 48h in low serum. (G) Average density profile of INTS11 across 76 EGF-responsive genes. The entire gene body is profiled, with the addition of 1kb upstream of the transcription start site (TSS) and 3kb downstream of the transcription end site (TES). Density profiles are normalized to sequencing depth. (See also Figure S1 and Table S1)

Since the RNAPII pause release mechanism is deemed to be unique to multi-cellular organisms (Kwak and Lis, 2013), we reasoned that the RNAPII-associated factor promoting the escape of RNAPII following EGF induction should display an evolutionary profile consistent with the appearance of RNAPII pause release mechanism in metazoans. Moreover, we envisioned that in order to induce signal-dependent release of RNAPII from the paused state, such a factor should be recruited to paused promoters following transcriptional induction. Based upon these criteria, we asked whether the Integrator, which is a metazoan-specific RNAPII-associated multi-protein complex, might fulfill the role of the RNAPII pause release factor following EGF transcriptional activation. We had previously characterized Integrator as a multi-subunit complex that was required for the maturation of the small nuclear RNA genes (Baillat et al., 2005) but the number of studies describing the wide array of developmental defects and diseased states upon its dysfunction suggest its scope may extend beyond the biosynthesis of snRNA (Han et al., 2006; Kapp et al., 2013; Tao et al., 2009).

While our initially developed antibodies against Integrator subunits detected its presence at UsnRNA genes, which are present in multiple copies within the genome, they were incapable of robust immunoprecipitation at non-repetitive mRNA encoding genes (Baillat et al., 2005). We screened a large number of commercially available antibodies against Integrator subunits to arrive at a set that displayed a greater recovery of immunoprecipitated chromatin. We used antibodies against three subunits of the complex (INTS11, INTS1 and INTS9) to determine the recruitment of Integrator to three candidate IEGs following stimulation with EGF using ChIP followed by real-time polymerase chain reaction (PCR). While, there was a small but detectable amount of Integrator at transcriptional start sites (TSS) prior to EGF induction, there was a robust increase in Integrator occupancy following EGF stimulation (Figure 1E). Integrator occupancy decreased 40 to 60 minutes after EGF induction, in accordance with the natural decay in the wave of transcription that is typical of IEGs (Figure 1E) (Amit et al., 2007). Unexpectedly, we also observed an increased occupancy of Integrator at the 3′-end of these three IEGs, suggesting the association of at least a subset of Integrator subunits with elongating RNAPII following transcriptional activation (Figure 1E).

To examine the detailed profile of Integrator prior to and following transcriptional activation, we performed ChIP-Seq using antibodies against INTS11. Interestingly, analysis of Integrator chromatin residence in cycling HeLa cells indicated its association with transcriptionally active genes mirroring that of RNAPII (Figure S1A). Moreover, as shown in Figures 1F and 1G, prior to stimulation with EGF in cells staved of serum, there was small amount of Integrator at the TSS of EGF-responsive genes. The addition of EGF resulted in increased occupancy of Integrator at the TSS and throughout the body of the EGF-responsive genes (Figure 1F and G). Analysis of 76 randomly chosen control genes non-responsive to EGF indicated that while EGF stimulation resulted in a small increase in Integrator at 5′-end of these genes, Integrator occupancy was not extended to the body of these genes (Figure S1B). These results demonstrate the stimulus-dependent recruitment of the Integrator complex to IEGs displaying paused RNAPII and the continued association of Integrator with productive RNAPII elongation complexes following pause release.

Integrator is critical for transcriptional activation of IEGs

We reasoned that if Integrator is required for the release of paused RNAPII into a productive elongation, its depletion should impact transcriptional activation induced by EGF treatment. Initially, we assessed the EGF responsiveness following Integrator depletion using microarrays. Depletion of INTS1 or INTS11 resulted in the loss of EGF responsiveness for nearly all of the 103 top EGF responsive genes on the array (corresponding to a minimum of 0.5 log₂ fold induction in the control sample, Figure 2A, B, Figure S2A and B and Table S1).

Figure 2. Integrator is essential for EGF responsiveness.

(A) Heatmap of the 116 top microarray probes (corresponding to 103 unique genes) upregulated by EGF in normal conditions (CTRL shRNA, t=30min over t=0). The color scale represents the modified log2 ratio (“sweep” function R, scaled by row) between the induced and the basal state. The color variation accounts for the difference of induction across the 2 conditions (Red=augmented induction, Green=decreased induction). The average of three independent experiments is shown. (B) Boxplot of the fold activation (expressed as log2) of the 116 most responsive microarray probes after EGF induction (*, p<0.01; Wilcoxon test). (C) EGF responsiveness at NR4A1, JUN, FOS and CCNL1 is dramatically decreased in the absence of INTS11. RNA-Seq analysis was performed in a cell clone expressing tet-inducible shRNAs targeting INTS11, doxycycline (DOX) was added to the culture medium for 72h to deplete INTS11 protein level. RNA-Seq tracks are aligned to the UCSC hg19 human genome. (D) Expression analysis of 76 EGF-responsive genes in a HeLa inducible shRNA clone by RNA-seq. A robust EGF-mediated activation of EGF genes (mean: 4.13 and 5.25, p=0.005) is impaired in the absence of INTS11 (mean: 3.95 and 4.52, p=0.15). The Boxplot represents the distribution of log2(FPKM) values for the top 76 genes induced by EGF in the control (log2FoldChange>0.4) and occupied by Integrator (see Fig. 1G). (E) Average density of RNA-Seq reads at 76 EGF responsive genes. Data were normalized to sequencing depth. (See also Figure S2 and Table S1)

We had previously failed to see changes in FOS expression following Integrator depletion (Baillat et al., 2005). This may have resulted from inadequate depletion of Integrator subunits or inefficient induction of the FOS gene. To validate the microarray results, we performed strand-specific RNA-Seq experiments measuring the activation of EGF-responsive genes 20 minutes following addition of EGF. To conduct these RNA-Seq experiments, we generated HeLa stable cell lines harboring doxycycline (DOX) inducible shRNA constructs (Figure S2C). Importantly, while depletion of Integrator resulted in a defect in UsnRNA processing as manifested by increased primary U2 transcripts after INTS11 or INTS1 depletion, we did not detect changes in the steady state levels of UsnRNAs during the time course of the experiments (Figure S2D). Moreover, we did not observe any defect in the downstream steps of EGF signaling such as ELK1 phosphorylation following Integrator depletion (Figure S2E). Finally, depletion of Integrator did not affect the cellular concentration of any of the RNAPII phosphorylation states (Figure S2F). Overall, these results indicate that the concentrations of UsnRNAs, the primary members of the EGF response pathway, and RNAPII levels/modifications are not significantly changed during the time course of the experiments.

Importantly, analysis of RNA-Seq data revealed that nearly all EGF responsive genes substantially diminished their transcriptional activation following Integrator depletion. This is clearly evident by examining RNA-seq reads corresponding to the steady state levels of messenger RNA for four representative IEGs genes (Figure 2C). Indeed, following Integrator depletion, EGF can no longer significantly activate the EGF-responsive genes (76 targets, induced by >1.4 fold in the control cell line) as shown in the box plot and the RNA-seq read density distribution of all 76 genes (Figure 2D and E). Taken together, these results demonstrate a critical role for the Integrator complex in EGF responsiveness.

Integrator regulates initiation and release from pausing following transactivation

To determine the mechanism of action of the Integrator complex, we performed global run-on sequencing (GRO-Seq) as well as RNAPII ChIP-seq following Integrator depletion to measure nascent RNA and RNAPII occupancy during EGF induction. Integrator depletion resulted in a profound decrease in both GRO-seq and RNAPII ChIP-seq signals corresponding to the body and the 3′-end of most EGF-responsive genes (Figure 3A), reflective of a role for Integrator in productive transcriptional elongation. This was also seen in the depth normalized read counts of the 76 EGF-responsive genes, where there was a strong reduction in read counts in the body of the EGF-responsive genes (Figure 3B and C, left). In contrast, we did not find any difference in the read counts of 76 control genes non-responsive to EGF (Figure 3B and C, right).

Figure 3. Integrator regulates initiation as well as pause release following EGF activation.

(A) Global Run-On (GRO-Seq) and RNAPII ChIP-Seq data were obtained from cells expressing doxycycline (DOX) -inducible shRNAs against INTS11, after EGF stimulation. JUN, CCNL1, and FOS profiles show a dramatic decrease in the amount of nascent RNA reads mapping to the gene body and 3′ end, whilst the peaks of RNA at the 5′ is proportionally increased. NR4A1 shows a moderate decrease across the gene body but a consistent accumulation of reads at the TSS. Concomitantly, INTS11 depletion dramatically reduces RNAPII occupancy at the body of all genes. All tracks represent read density normalized to sequencing depth. (B) Average density of GRO-Seq reads at 76 EGF responsive genes. Mean densities were normalized to sequencing depth. The entire gene locus is displayed with additional 1kb upstream the TSS and 3kb downstream the TES. The right panel shows a control analysis performed on 76 transcriptionally active genes non responsive to EGF (Table S1). These genes were randomly chosen among the most active genes in HeLa cells according to RNAPII occupancy. (C) Average density of RNAPII ChIP-Seq reads at 76 EGF responsive genes. Mean densities for all genes were normalized to sequencing depth. The right panel shows a control profile of RNAPII at 76 control genes. (D) Residual occupancy of RNAPII after INTS11 depletion in EGF stimulated cells. The percentage is calculated from the ratio of the average read distribution with or without DOX (the average gene locus was divided in 35 bins from −500 upstream the TSS to 1kb downstream the TES). Depletion of Integrator affects elongating RNAPII in the gene body and 3′ end to a greater extent than initiating RNAPII. (E) Traveling Ratio of RNAPII in the presence or absence of INTS11. TR increases at nearly all EGF-responsive genes in the absence of INTS11, indicating accumulation of non-processive paused RNAPII. (See also Figure S3)

Moreover, Integrator depletion also resulted in decreased transcriptional initiation as seen in the reduction of Gro-seq and RNAPII ChIP-seq depth normalized read counts on the 5′-end of EGF-responsive genes (Figure 3B and C). We further analyzed RNAPII occupancy on the 5′- and 3′-end of FOS, JUN and NR4A1 using ChIP followed by real-time PCR. These results indicated that while Integrator depletion resulted in decreased levels of RNAPII on the 5′-end of genes, reflective of a defect in transcriptional initiation, the RNAPII occupancy displayed a greater reduction on the 3′-end of these three EGF-responsive genes (Figure S3A and B). Indeed, the defect in transcriptional elongation following depletion of Integrator was also evident following analysis of elongating serine 2 phosphorylated form of RNAPII (Figure S3C). This is also seen following the analysis of residual RNAPII chromatin residence following depletion of Integrator reflecting a larger reduction in the elongating RNAPII (Figure 3D). Finally, measurement of the RNAPII travel ratio confirmed a decrease in RNAPII release from pausing specific to EGF responsive genes (Figure 3E and S3D). Taken together, our combined results of GRO-Seq and RNAPII profiling of EGF-responsive genes support a role for Integrator in promoting pause release leading to productive transcriptional elongation following EGF induction. Moreover, we also find a role for Integrator in recruitment of RNAPII to the 5′-end of EGF-responsive genes during transcriptional activation. This was more evident at some of the EGF-responsive genes such as NR4A1 and JUN (Figure S3A and B).

Integrator is critical for the recruitment of super elongation complex following EGF induction

The current models for the mechanism leading to the release of paused RNAPII stipulate a requirement for the recruitment of the P-TEFb complex (CDK9 and Cyclin T) following transcriptional induction (Kwak and Lis, 2013). Recent biochemical studies have identified a large transcriptional elongation complex termed super elongation complex (SEC) containing multiple subunits including ELL2 and AFF4, in addition to the most active version of the elongation factor P-TEFb (CDK9/Cyclin T) (Lin et al., 2010; Luo et al., 2012). Therefore, we next asked whether the block in the release of paused RNAPII following Integrator depletion could be the result of defective/impaired recruitment of SEC. We performed ChIP-Seq using antibodies against ELL2 and AFF4 as a mark for active P-TEFb complex to assess the contribution of Integrator to SEC recruitment. Depletion of Integrator nearly abolished the stimulus-dependent recruitment of ELL2 and AFF4 to IEGs (Figure 4A and Figure S4A). This was not only reflected in the ELL2 and AFF4 profiles of individual genes but also on the average profile of all 76 EGF-responsive genes (Figure 4B). Integrator depletion did not change SEC recruitment of 76 control transcriptionally active genes unresponsive to EGF (Figure S4B).

Figure 4. Integrator is required for SEC recruitment.

(A) Recruitment of AFF4 and ELL2 components of the SEC complex is severely impaired in the absence of Integrator. ChIP-Seq data were obtained from a cell clone expressing tet-inducible shRNAs targeting INTS11, before and after EGF stimulation. JUN and FOS profiles show a dramatic decrease in the amount of SEC recruited (see Fig. S4A for NR4A1 and CCNL1 profiles). Y axis represents read density normalized to sequencing depth. (B) Average density of AFF4 and ELL2 ChIP-Seq reads at 76 EGF responsive genes. Distribution are shown for INTS11 knock-down and its control, before (dashed lines) and after (solid lines) induction with EGF. Density profiles were normalized to total read number. (See also Figure S4)

We confirmed these results by depleting Integrator and measuring the occupancy of the CDK9 component of SEC following EGF induction using real-time PCR. While depletion of Integrator had no effect on global levels of CDK9 (Figure S4C), its depletion abrogated the EGF-induced recruitment of CDK9 to the FOS, JUN and NR4A1 promoters as measured by ChIP, similar to that seen with ELL2 and AFF4 (Figure 5A).

Figure 5. Integrator functionally and physically interacts with SEC.

(A) INTS11 knockdown prevents CDK9 recruitment at IEG promoters in response to EGF stimulation. CDK9 occupancy was measured by qChIP after INTS11 knockdown and EGF induction. Data are the average of three independent experiments (*, p<0.15; **, p<0.05, t-test). (B) Co-immunoprecipitation of overexpressed Myc-CDK9 and Flag-Integrator subunits in HEK293T cells shows CDK9 interaction with INTS1, INTS4 and INTS11. (C) Endogenous Integrator and CDK9 co-precipitate in HEK293T nuclear extracts, immunoprecipitations were performed with INTS1, INTS4, INTS11 and two different CDK9 antibodies. The negative elongation factor, NELFA, was used as a negative control. (D) Co-immunoprecipitation of SEC and Integrator using antibodies against endogenous CDK9 and AFF4. During the course of the IP, nuclear extract of HEK293T was incubated with 50 ug/ml Ethidium Bromide to disrupt indirect protein interactions mediated by DNA (upper panels) and with RQ1 DNase or RNase T1/RNase H to disrupt indirect protein interactions (lower panel).

These results prompted us to ask whether Integrator and components of the SEC physically interact. We found a physical association between the ectopically transfected Flag-Integrator subunits and Myc-CDK9 (Figure 5B). Furthermore, reciprocal co-immunoprecipitation demonstrated a robust association between endogenous Integrator and the CDK9 component of the SEC complexes (Figure 5C). Finally, we asked whether Integrator and SEC association is mediated through nucleic acids. Treatment of the immunoprecipitation samples with 50 ug/ml of ethidium bromide (Figure 5D. upper panel), RNase, or DNase (Figure 5D, lower plane) did not affect the association between Integrator and SEC following immunoprecipitation with CDK9 or AFF4 antibodies, confirming that the interaction is not nucleic acid-dependent. Taken together, our results provide support for the Integrator complex recruitment of SEC to the promoter of EGF responsive genes leading to pause release and increased transcriptional elongation.

The Integrator function in pause release is conserved in Drosophila

Transcriptional response of the HSP70 gene to heat shock in Drosophila represents one of the first examples of RNAPII pause release mechanism in animals (Rasmussen and Lis, 1993). Since most of the Integrator subunits are conserved in metazoans (Peart et al., 2013), we asked whether Integrator plays a role in HSP70 responsiveness to heat shock in Drosophila. We first assessed RNAPII and Integrator (using antibodies against dINTS9 and dINTS12) occupancy by ChIP using primers to the 5′-end, body and the 3′-end of the gene prior to and following heat shock (Figure 6A). Consistent with observations in mammalian cells (Figure 1, and Figure S1), Integrator displayed a similar profile to that of RNAPII, occupying the promoter and traveling through the body of the HSP70 gene following heat shock treatment (Figure 6B-D). Recruitment of Integrator following heat shock induction suggested a role for this complex in HSP70 transcriptional activation in fly. To address this possibility, we depleted Integrator by means of double-stranded RNAs (dsRNAs) against either dINTS1, dINTS9, dINTS11 or dINTS12 (Figure 6E). DsRNAs against LacZ was used as control (Figure 6E). Depletion of Integrator resulted in a substantial diminution of HSP70 responsiveness to heat shock (Figure 6F). These results not only indicate that Integrator function in transcriptional activation is conserved in Drosophila but also extends its scope of function in modulation of pause release in this organism.

Figure 6. Integrator is required for Heat Shock activation in Drosophila.

(A) Position of the qPCR amplicons along the Drosophila melanogaster HSP70Aa gene. (B) qChIP analysis of RNA Polymerase II (unmodified CTD) localization on the HSP70 gene before and after heat shock (S2 cells, 5 min at 37C). (C) qChIP analysis of IntS9 localization on the HSP70 gene before and after heat shock (S2 cells, 5 min at 37C). (D) qChIP analysis of IntS12 localization on the HSP70 gene before and after heat shock (S2 cells, 5 min at 37C). (E) Immunoblot analysis of Integrator subunits after treatment of S2 cells with dsRNA against each Integrator subunit, dsRNA targeting the E. coli LacZ gene is used as a negative control. (F) Induction of HSP70 transcription by 20 min heat shock at 37C after knockdown of different Integrator subunits and LacZ as a control. The fold induction is compared with room temperature expression levels after normalization using the RPS17 gene, a control housekeeping gene (average of three independent experiments). (See also Figure S5)

Discussion

The novelty of our work lies in the following: first, we show that nearly all EGF-responsive genes display release from pausing following EGF induction. Second, we find that Integrator is recruited to protein-coding genes in a signal dependent manner and displays a chromatin occupancy pattern reflective of its role in transcriptional elongation. Third, we demonstrate that loss of Integrator results in the abrogation of IEGs responsiveness to EGF induction, consistent with a key role for this complex in regulation of transcriptional activation. Fourth, through analysis of nascent RNA transcription and RNAPII profiling, we demonstrate that Integrator plays a key role in both transcriptional initiation and pause release. Fifth, we show that Integrator physically associates with the SEC complex and it is required for its recruitment following transcriptional activation. Finally, analysis of heat shock response in HSP70 gene extends the role for Integrator in transcriptional activation and pause release to Drosophila melanogaster.

Transcriptional paused release was initially reported in mammals using the c-Fos gene as a model system (Plet et al., 1995). We used the IEGs responsive to EGF to determine whether such a block in RNAPII productive elongation is a common feature of EGF-responsive genes prior to their activation. EGF signaling offers a convenient model to study pause release mechanism following transcriptional activation since serum starvation constrains the IEGs revealing paused RNAPII at their transcriptional start sites (Figure 1A). Addition of EGF will then result in the activation of a signaling cascade that culminates in the release of such pausing allowing RNAPII entry into a productive transcriptional elongation (Figure 7). This was evidenced by the change in RNAPII traveling ratio following EGF induction (Figure 1D), which has been previously used as a faithful measure for pause release in mammalian cells (Rahl et al., 2010). It is important to note that the EGF-induced activation also results in increased recruitment of RNAPII (Figure 1A and B) and therefore transcription is also impacted at the initiation step.

Figure 7. Integrator’s role in EGF-mediated gene activation.

We propose a model for the role of the Integrator complex in elongation and pause release. During serum starvation, an immediate early gene reduces its transcriptional activity as paused polymerase is accumulated at the TSS. Upon EGF stimulation, a signaling cascade results in a sequence-specific binding of a combination of transcription factors at the promoter of IEGs. Following activation, Integrator is directed to EGF target genes through the C-terminal domain of RPB1 resulting in the recruitment of the SEC complex leading to eviction of negative elongation factors and inducing of productive transcriptional elongation.

It has been proposed that the initiating RNAPII encounters an impediment to transcriptional elongation somewhere between nucleotides 20 to 60 in mammalian cells resulting in RNAPII stalling (Kwak and Lis, 2013). Studies in mammalian cells and Drosophila melanogaster have revealed that such RNAPII pausing may result from the action of two transcriptional regulatory factors, NELF (Negative Elongation Factor) and DSIF (DRB Sensitivity-Inducing Factor) providing a physical barrier to elongating RNAPII (Martinez-Rucobo et al., 2011; Missra and Gilmour, 2010). Furthermore, a combination of sequence composition and chromatin architecture close to transcriptional start sites, may further impose an obstacle against productive elongation (Gilchrist et al., 2010; Izban and Luse, 1991; Mavrich et al., 2008). However, upon activation the recruitment of positive elongation factor P-TEFb was proposed to relieve such a paused state by phosphorylating NELF, DSIF and the serine 2 residue of the largest subunit of RNAPII resulting in displacement of the factors and the release of RNAPII into a productive elongation. In that context, we show that for nearly all IEGs responsive to EGF the Integrator complex plays a pivotal role in recruitment of P-TEFb. Indeed, P-TEFb resides in a large multi-protein complex termed super elongation complex (SEC), which physically and functionally associates with the Integrator complex. Our immunoprecipitation experiments revealed that only a fraction of Integrator and SEC complexes were physically associated and that such an association may be triggered through a signaling cascade. Interestingly, in the absence of Integrator, SEC is no longer recruited to IEGs following transcriptional activation resulting in the persistent state of pausing and abrogation of transcriptional activation (Figure 7).

The mechanisms governing transcriptional elongation are at the heart of transcriptional regulation in metazoans and recent estimates suggest that nearly half of mammalian genes are regulated through pause release (Kwak and Lis, 2013; Muse et al., 2007; Rahl et al., 2010; Zeitlinger et al., 2007). Our observation that Integrator plays a similar role in the activation of HSP70 gene following heat shock in Drosophila melanogaster extends its role in regulation of pause release in metazoans. Taken together, these observations point to Integrator complex as an evolutionary conserved RNAPII-associated complex that may play a critical role in the synchronous activation of gene expression during metazoan development.

We show that while Integrator occupied the IEGs in serum-starved cells, it was further recruited following EGF stimulation. RNAPII displayed a similar patter of chromatin residence prior to and following transcriptional activation. Interestingly, depletion of Integrator also diminished the recruitment of RNAPII to the 5′-end of many IEGs following EGF stimulation. These results indicate that Integrator also modulates transcriptional initiation following EGF induction. Importantly, detailed quantification of changes of RNAPII chromatin residence at 5′- versus 3′-end of FOS, JUN and NR4A1 indicated that Integrator depletion has a larger impact on the occupancy of RNAPII at 3′-end or the elongating RNAPII (Figure S3). Nevertheless, these results highlight an important contribution of Integrator to RNAPII recruitment during transcriptional activation and suggest the possibility of the signal transduction pathways targeting Integrator subunits for post-translational modification resulting in enhanced recruitment.

We initially defined Integrator as a stable multi-subunit complex that associates with the CTD of RNAPII and is involved in the biogenesis of UsnRNA genes. Integrator is composed of at least 14 subunits, most of which are yet to be fully characterized. The INTS11/INTS9 heterodimer endows Integrator with an endonuclease activity required for UsnRNA processing (Baillat et al., 2005) while INTS3 and INTS6 subunits are also components of other protein complexes that participate in DNA repair (Huang et al., 2009; Skaar et al., 2009). Our genome-wide analyses indicated a far broader involvement of Integrator in transcriptional regulation than was previously anticipated, extending its role to the activation of protein-coding genes. The exact role that different modules and subunits of Integrator play in such diverse biological processes will be an important aspect of understanding Integrator’s function. Placed at the crossroads of transcriptional elongation and RNA processing, Integrator is emerging as a multifaceted platform that coordinates and integrates different extracellular signals translating them into a proficient and productive transcriptional response in higher eukaryotes.

Experimental procedures

Genome-wide data

All High-throughput sequencing and Microarray data presented in this study are deposited at the Gene Expression Omnibus (GSE58255).

Cell culture

HeLa cells were grown in high-glucose DMEM, supplemented with 2 mM L-glutamine and 10% fetal bovine serum. Drosophila S2 cells were grown in Schneider medium supplemented with 10% FBS.

Antibodies

Antibodies against INTS1, INTS4, INTS11, Phospho-Ser2 CTD, phosphor-ser5 CTD and CDK9 (used for immunoprecipitation in Fig. 4c) were obtained from Bethyl Laboratories. Antibodies against RNAPII (N-20, rabbit polyclonal antibodies recognizing all forms of RNAPII), ELK-1, phospho-ELK1 and CDK9 (H169 used for ChIP, D7 for western blot and C20 for immunoprecipitation) were obtained from Santa Cruz. Additional polyclonal antibodies used for ChIP or ChIP-seq are raised against INTS9 (Bethyl lab.), RNAPII (N20, Santa-Cruz), AFF4 and ELL2 (Lin et al., 2010). Additional qChIP was performed as previously described (Baillat et al., 2012) with antibodies against INTS1, INTS9, RNAPII phospho-ser-2, phosphor-ser5 CTD (Bethyl lab.).

ChIP-seq

25-30 × 10⁶ asynchronously growing HeLa cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature, harvested and washed twice with 1× PBS. The pellet was resuspended in ChIP lysis buffer and chromatin was sheared to an average length of 200-400 bp, using a Bioruptor sonication device (Diagenode). The chromatin lysate was diluted with SDS-free ChIP lysis buffer and aliquoted into single IPs of 2,5×10⁶ cells each. A specific antibody or a total rabbit IgG control was added to the lysate along with Protein A magnetic beads (Invitrogen) and incubated at 4°C overnight. On day 2, beads were washed twice with Mixed Micelle Buffer, Buffer 500, LiCl/detergent wash and once with 1× TE. Finally, beads were resuspended in 1× TE containing 1% SDS and incubated at 65°C for 10 minutes to elute immunocomplexes. Elution was repeated, and the samples were further incubated overnight at 65°C to reverse cross-linking, along with the untreated input (2,5% of the starting material). After treatment with 0,5 mg/ml proteinase K for 3 hours, DNA was purified with Wizard SV Gel and PCR Clean-up system (Promega). DNA concentration was assessed with Quantit PicoGreen dsDNA kit (Invitrogen) and 5 to 10 ng were used to generate the sequencing libraries. DNA fragments of ~150-400 bp range were isolated by agarose gel purification, ligated to primers, and then subject to Solexa sequencing using manufacturer recommendations (Illumina, Inc.).

ChIP-seq data analysis

Reads were aligned to the human genome hg19 using bowtie2 (Langmead et al., 2009) (default parameters). Snapshots of raw ChIP-seq data presented throughout the figures were obtained as follows: BigWiggle files for every ChIP-Seq were generated using samtools, bedtools and RseQC (Wang et al., 2012), these tracks were then uploaded into the UCSC Genome Browser hg19.

GRO-seq

GRO-seq experiments were performed as previously described (Core et al., 2008; Wang et al., 2011). Briefly, HeLa cells were washed twice with ice-cold PBS before adding swelling buffer. Cells were swelled for 5 min on ice and then lysed in lysis buffer. Nuclei were washed twice with lysis buffer and resuspended in 100μl freezing buffer. An equal volume of reaction buffer was added and incubated for 7 min at 30°C for the nuclear run-on. Nuclear run-on RNA was extracted with TRIzol LS reagent (Invitrogen) following the manufacturer’s instructions and ethanol precipitated. Re-suspended NRO-RNA was treated with DNase (Ambion) for 30 min and hydrolyzed using fragmentation reagents (Ambion) for 13 min at 70°C. After purification through a Micro Bio-Spin p-30 column (Bio-Rad), T4 PNK (NEB) was used to repair the NRO-RNA ends. The Br-UTP-labeled NRO-RNA was then purified by anti-BrdU agarose beads (Santa Cruz Biotech) in binding buffer for 1h. Following elution, RNA was ethanol-precipitated prior to denaturing RNA and treating with poly(A)-polymerase (NEB) for 30 min at 37°C. cDNA synthesis was performed as described previously (Wang et al., 2011) using oNTI223 primer. The reaction was treated with 3 μl exonuclease I (Fermentas) for 15 min at 37°C, followed by 2 μl 1M NaOH for 20 min at 98°C, and neutralized with 1μl 2M HCl. cDNA was run on a 10% TBE-urea gel and products were excised and eluted from shredded gel pieces for 4h in TE + 0.1% Tween and precipitated in ethanol overnight. First-strand cDNA was circularized with CircLigase (Epicentre), denatured for 10 min at 80°C, and relinearized with APE I (NEB). Linearized DNA was amplified by PCR using Phusion Hot Start II Kit, according to manufacturer’s instructions. The oligonucleotide primers oNTI200 and oNTI201 were used for amplification. The PCR product was run on a 10% TBE gel and eluted as before. Libraries were sequenced on an Illumina hi-Seq2000.

Clustering and Heatmap analysis

ChIP-seq data were subjected to unbiased clustering, with respect to a list of unique RefSeq genes, using the seqMINER 1.3.3 package (Ye et al., 2011). We used Kmeans linear as the method of clustering, with the following parameters: left and right extension=1.5 kb, internal bins=160, flanking region bins=20 number of cluster=8. seqMINER was also used to generate all the heatmaps and extract read densities at EGF and control genes. Mean density profiles were then generated in R and normalized to sequencing depth.

RNA-Seq data analysis

Total RNA was extracted and treated with DNAse using RNeasy mini columns (QIagen). RNA was depleted of ribosomal RNA using RiboZero (Epicentre) and processed with the ScriptSeqv2 kit along with ScriptSeq Index PCR primers (Epicentre) to generate a strand specific library of total RNA. Reads were aligned to the human genome using bowtie2 (Langmead and Salzberg, 2012) (default parameters). FPKM for each gene was calculated using cufflink (Trapnell et al., 2010).

Traveling Ratio

RNAPII distribution across the gene body was estimated using the Traveling Ratio as an estimate of pause release. Data were generated as originally described (Rahl et al., 2010). Briefly, ChIP-seq read density at the TSS (−30bp to +300 bp) was divided by the read density over the rest of the gene body, plus an additional 1kb beyond the TES. Log10(ratio) of genes (EGF or Control) were calculated using all different isoforms available in the Hg19 RefSeq Annotation Table.

Heat Shock Treatment

S2 cells were grown in Schneider medium supplemented with 10% FBS. To induce heat shock, one volume of medium pre-heated to 48C was added to the medium and gently mixed immediately to raise the temperature to 37C. The cells were then immediately transferred to a 37C incubator for the indicated time. After heat shock, another volume of medium pre-chilled to 4C was added and gently mixed immediately to bring the temperature back to room temperature. The cells were immediately fixed (for ChIP) or lyzed with trizol (for RT-qPCR).

ChIP analysis of HSP70

S2 cells were fixed with 1% HCHO at rt for 7 min. The reaction was quenched with 125 mM glycine. After lysis, cells were sonicated at 4C to obtain 500 bp DNA fragments. Lysates of 7,5 million cells were diluted to 1ml with IP buffer, pre-cleared with Protein-A beads and incubated overnight at 4C with 1ug of antibody. The next day, lysates were immunoprecipitated with protein A beads for 2 hours at 4C and washed once with low salt buffer, thrice with high salt buffer, once with LiCl buffer and twice with 1× TE. Immunocomplexes were eluted and de-crosslinked at 65C overnight. DNA was extracted by phenol/chloroform and ethanol precipitated. DNA was resuspended in 100uL and 2uL were used for each PCR reaction.

RNAi

S2 cells were grown in serum free medium (SF900-II, life technologies). Templates for in vitro dsRNA transcription were generated by PCR using oligonucleotides containing T7 RNA polymerase promoter sequences at their 5′ ends (see SI for sequences). dsRNAs were generated by in vitro transcription with T7 RNA polymerase using standard methods. After phenol/chloroform extraction and ethanol precipitation, dsRNAs were added to the medium at a concentration of 10 ug/mL 24h, 48h and 72h before heat shock.

shRNA

pSUPER.retro.puro (oligoengine) constructs against INTS1 and INTS11 and a non targeting control (see SI for sequences) were transfected in HeLa cells using MetafectenePro (Biontex) as a carrier. After 24 hours, cells were selected with 2.5 μg/mL puromycin for 72 hours and serum starved for 24 hours in 0.5% serum. RNA was extracted before and after 30 min treatment with 100 ng/mL rEGF (Invitrogen).

The same shRNA sequences targeting INTS1 and INTS11 were cloned in a Tet-pLKO-puro (addgene). Lentiviral particles were produced in HEK293T cells and used to infect HeLa cells. After 48 hours selection in 2.5 μg/mL puromycin cells were diluted, single clones isolated, and screened by western blot for efficient knockdown. Knockdowns were induced by addition of 1μg/mL doxycycline in the culture medium for up to 96 hours. For EGF inductions, after 72 hours the medium was replaced by DMEM supplemented with 0.5% serum and 1μg/mL doxycycline for 24 hours followed by 100 ng/mL EGF treatment for 30 min.

HIGHLIGHTS.

• -Integrator is recruited to protein coding genes.
• -Loss of Integrator abrogates Immediate Early Gene responsiveness to EGF.
• -Integrator plays a role in transcription initiation and pause release.
• -Integrator physically associates with the SEC complex.