The regulation of gene expression programmes is essential for the generation of diverse cell types during development and for adaptation to environmental signals. RNA polymerase II (RNAPII) transcribes genetic information and coordinates the recruitment of accessory proteins that are responsible for the establishment of active chromatin states and transcript maturation. RNAPII is post‐translationally modified at active genes during transcription initiation, elongation and termination, and thereby recruits specific histone and RNA modifiers. RNAPII complexes are also located at silent genes in promoter‐proximal paused configurations that provide dynamic transcriptional regulation downstream from initiation. In embryonic stem cells, silent developmental regulator genes that are repressed by Polycomb are associated with a form of RNAPII that can elongate through coding regions but that lacks the post‐translational modifications that are important for coupling RNA synthesis to co‐transcriptional maturation. Here, we discuss the mechanisms through which the transcription of silent genes might be dissociated from productive expression, and the sophisticated interplay between the transcriptional machinery, Polycomb repression and RNA processing.
See Glossary for abbreviations used in this article.
RNA polymerase II (RNAPII) transcribes all protein‐coding genes and many non‐coding RNAs. RPB1, the largest subunit of RNAPII, contains an unusual CTD that comprises repeats of the heptapeptide consensus sequence N‐Tyr 1‐Ser 2‐Pro 3‐Thr 4‐Ser 5‐Pro 6‐Ser 7‐C, of which there are 52 copies in mammals (Corden et al, 1985). All of the amino acids in these repeats are potential targets for post‐translational modification—such as phosphorylation, glycosylation and cis–trans isomerization—and these modifications dictate the functional state of the polymerase. In addition, lysine residues in the several non‐consensus CTD repeats that are found in mammals can act as targets for ubiquitination (Li et al, 2007). Finally, in response to DNA damage or drug‐induced RNAPII stalling, RNAPII is ubiquitinated outside the CTD, which is crucial for its proteasomal degradation (Bregman et al, 1996; Somesh et al, 2005).
Although the roles and extent of these RNAPII CTD modifications are far from understood, their vast combinatorial potential generates a complex regulatory code (Buratowski, 2003). CTD phosphorylation during the transcription cycle helps to recruit factors that modulate chromatin states and RNA processing, thereby integrating transcription with genome architecture and RNA stability (Phatnani & Greenleaf, 2006). RNAPII is also present across silent regions of the genome, including inactive genes (Guenther et al, 2007; Hargreaves et al, 2009; Muse et al, 2007), enhancer regions of silent genes (reviewed by Szutorisz et al, 2005), and Polycomb‐repressed chromatin (Chopra et al, 2009; Dellino et al, 2004; Stock et al, 2007). Its widespread presence across the genome suggests that RNAPII regulation is crucial for controlling the status of gene expression and has significant consequences for cell response pathways.
RNAPII regulation at active genes
RNAPII is recruited to gene promoters in a hypo‐phosphorylated state (Fig 1). Its escape from the promoter requires TFIIH‐mediated phosphorylation of Ser 5 (Ser 5P; Komarnitsky et al, 2000), which recruits the RNA capping machinery to modify the 5' end of nascent RNAs (Ho & Shuman, 1999) and, in yeast, the Set1 HMT to methylate histone H3K4 (Fig 2A; Ng et al, 2003). During the early stages of the transcription cycle, instability of the RNAPII complex can cause transcription to be aborted after only a few nucleotides (Fig 1). After promoter escape, RNAPII proceeds to intrinsic pausing sites where it can be halted by negative factors. The onset of productive elongation requires P‐TEFb, which phosphorylates Ser 2 residues of the RNAPII CTD to produce stable elongation complexes (Fig 1B; Marshall et al, 1996). In both yeast and mammals, Ser 2P integrates productive elongation with chromatin remodelling through the recruitment of H3K36 HMTs, which promote chromatin that is compatible with elongation (Fig 2A; Krogan et al, 2003; Li et al, 2005). Ser 2P also links RNA synthesis with co‐transcriptional RNA processing by helping to recruit splicing and polyadenylation factors (Fig 2A; Proudfoot et al, 2002).
RNAPII termination involves the release of RNA transcripts and the dissociation of transcription complexes from the DNA template, and is functionally correlated with transcript polyadenylation (Richard & Manley, 2009). CTD phosphatases are required to recycle RNAPII and produce its hypo‐phosphorylated, initiation‐competent form (Fig 1). In yeast, Fcp1 removes phosphates from Ser 2 residues (Cho et al, 2001), whereas Ssu72 (Krishnamurthy et al, 2004) and Rtr1 (Mosley et al, 2009) target Ser 5P. Small CTD phosphatases (SCP1–3), recently characterized in mammals, cooperate with the REST chromatin remodelling complex, conferring the potential to regulate RNAPII at specific genes (Yeo et al, 2005).
The post‐translational modification of CTD residues other than Ser 2 and Ser 5 is less understood. Phosphorylated Ser 7 residues have been identified at promoter and coding regions of actively transcribing genes in a human tumour cell line (Chapman et al, 2007). Ser 7P is essential for the transcription of small nuclear RNAs (Egloff et al, 2007), but its genome‐wide influence on protein‐coding gene expression requires clarification. Similar to Ser 5, Ser 7 is phosphorylated by TFIIH, and so could also be required for the early stages of transcription (Akhtar et al, 2009). Interestingly, Ser 7P might influence the extent of Ser 5 phosphorylation (Chapman et al, 2007), thereby helping to modulate the degree of transcription‐coupled chromatin remodelling or RNA processing. An additional CTD regulator is the phospho‐Ser/Thr‐Pro‐directed isomerase Pin1, which catalyses the conversion between proline cis–trans configurations, and therefore alters CTD folding (Lu & Zhou, 2007). Mammalian Pin1 regulates RNAPII phosphorylation and activity during the early stages of the transcription cycle and might have a role in RNA processing (Xu & Manley, 2007).
Promoter‐proximal pausing of RNAPII
The promoter‐proximal regions of many inactive genes are associated with Ser 5P RNAPII in the absence of detectable expression. This phenomenon is known as promoter‐proximal pausing (Fig 1A), a state in which RNAPII is transcriptionally engaged but becomes stalled by negative elongation factors (Core & Lis, 2008). Promoter‐proximal pausing of RNAPII was first shown to occur at the Drosophila melanogaster heat‐shock gene hsp70 (Gilmour & Lis, 1986), where the recruitment of RNAPII before gene activation allows transcription to commence rapidly on heat shock (Boehm et al, 2003).
RNAPII has been detected in human cells at inactive genes such as the proto‐oncogene c‐MYC (Eick et al, 1987) and the major histocompatibility complex class II DRA gene, where transcription can be rapidly induced by IFN‐γ (Spilianakis et al, 2003). Similarly, basal levels of p53 in human cancer cells allow the recruitment of RNAPII to the non‐induced p21 gene before its conversion to an elongating complex after activation (Espinosa et al, 2003).
Pausing affects initiating RNAPII molecules to different extents, either preventing all but very infrequent full‐length transcription (Fig 1A), or regulating the recurrent release of RNAPII into productive elongation at active genes (Fig 1B; Core & Lis, 2008). Recent genome‐wide profiling of RNAPII in human fibroblasts and ES cells, as well as in Drosophila cells, has shown the widespread presence of low levels of RNAPII bound at inactive gene promoters (Guenther et al, 2007; Kim et al, 2005; Muse et al, 2007; Zeitlinger et al, 2007). These studies do not provide evidence for the transcriptional competence of bound RNAPII, although here we define paused genes by their association with promoter‐proximal RNAPII in the absence of overt transcription. Global sequencing of nascent RNAs that are labelled by Br‐UTP incorporation in nuclei purified from human cells, has revealed transcriptionally engaged RNAPII at virtually all genes (Core et al, 2008). The sensitivity of this technology allowed the detection of full‐length RNAs from paused genes, suggesting that a low level of RNAPII escape occurs from promoter‐proximal regions. Such findings indicate that regulating the release of RNAPII from promoter‐proximal pausing is a universally important control point that primes genes for future activation, for example by promoting open chromatin states (Gilchrist et al, 2008; Lee et al, 2008).
Pausing factors such as NELF and DSIF prevent the exit of RNAPII from promoter‐proximal regions into productive elongation (Cheng & Price, 2007). The negative actions of NELF and DSIF are reverted by P‐TEFb‐dependent phosphorylation, which induces the dissociation of NELF and the activation of DSIF as a positive elongation factor (Fujinaga et al, 2004; Yamada et al, 2006). The presence of abundant NELF at active gene promoters in Drosophila (Lee et al, 2008) supports the view that pausing is an important checkpoint that regulates RNA expression levels at active genes (Fig 1B).
Poised RNAPII at Polycomb‐repressed genes
Polycomb‐mediated repression is crucial for maintaining the silent state of genes in many cell types (Schwartz & Pirrotta, 2008). Two major Polycomb repressive complexes, PRC1 and PRC2, promote gene silencing through the modification of chromatin structure, by inducing histone H2AK119 monoubiquitination and H3K27 trimethylation, respectively (Cao et al, 2002; Wang et al, 2004). PRCs can also induce chromatin condensation indirectly by influencing DNA methylation (Sakamoto et al, 2007; Vire et al, 2006). However, PRC‐mediated repression does not prevent the transcriptional apparatus from accessing promoters (Breiling et al, 2001; Chopra et al, 2009; Dellino et al, 2004; Stock et al, 2007), suggesting that silencing could involve a direct interference with the transcriptional machinery, in addition to chromatin remodelling.
The binding of general transcription factors, including TBP, to Polycomb‐repressed homeotic gene promoters has been observed in Drosophila cells (Breiling et al, 2001). Moreover, Drosophila Polycomb proteins associate with general transcription factors in vitro (Saurin et al, 2001). In transgenic Drosophila embryos, in which a Polycomb response element was placed upstream from Hsp26, Polycomb proteins did not prevent the recruitment of the transcriptional machinery to the promoter, but instead repressed the onset of transcription (Dellino et al, 2004). Furthermore, PRC‐bound chromatin can maintain an open conformation in Drosophila cells (Mito et al, 2007) and transcriptional activity has been detected at PRC‐bound genes in several systems (Bracken et al, 2006; Breiling et al, 2004; Chopra et al, 2009; Stock et al, 2007; Tolhuis et al, 2006), including at the Polycomb genes (Schwartz et al, 2006). Therefore, the presence of RNAPII at Polycomb‐silenced genes could be important to modulate their level of expression or to allow later expression. In this respect, the importance of RNAPII regulation at PRC‐repressed genes has recently been shown in mouse ES cells (Stock et al, 2007).
Developmental regulator genes are maintained in a silent state in ES cells by PRCs to prevent the activation of differentiation programmes (Boyer et al, 2006; Endoh et al, 2008). Surprisingly, many developmental genes are marked by histone modifications that are characteristic of the active state (H3K4 trimethylation and H3K9 acetylation), as well as repressive modifications that are catalysed by Polycomb proteins (H3K27 trimethylation and H2AK119 monoubiquitination; Sidebar A; Azuara et al, 2006; Bernstein et al, 2006; Stock et al, 2007). Despite PRC repression, the promoters of genes with bivalent chromatin marks are bound by similar amounts of RNAPII as active genes, and RNAPII can be detected at least 8 kb downstream from transcription start sites (Fig 1C; Stock et al, 2007). Here, we refer to the specific RNAPII configuration found in ES cells at PRC‐silenced genes as ‘poised’ because of its important role in potentiating genes for activation during development, but we anticipate that similar findings will be described in other cell types.
Sidebar A | RNA polymerase II and histone modification profiles across genes in paused, active and poised states
Distinct RNAPII complexes at paused, active and poised genes can be identified by mapping the binding of antibodies that recognize specific modifications along transcription units by chromatin immunoprecipitation (ChIP). ChIP is performed on cell populations, and so these profiles capture RNAPII distribution simultaneously across the gene in many cells. The profiles below are thought to be representative of mammalian genes, although many variations on these will be observed across the genome and in different cell types.
Paused genes (A) contain low levels of Ser 5P limited to promoter‐proximal regions. (B) Actively transcribing genes are associated with Ser 5P enrichment peaking at the promoter and extending through the gene, whereas (C) at poised PRC‐repressed genes, Ser 5P is highly enriched within approximately 8 kb of the TSS. Consistent with a role in processive elongation, Ser 2P is present throughout the coding regions of active genes, but is absent from paused and poised complexes. Antibody 8WG16, which recognizes non‐phosphorylated Ser 2 residues, binds to RNAPII at promoters and into coding regions of active genes, but not towards the 3′ end of genes. 8WG16 enrichment is restrained to promoter‐proximal regions of paused genes, whereas very low levels, or a lack of binding, characterize poised genes.
Histone modification profiles also distinguish each transcriptional state: H3K4 trimethylation marks paused, active and poised genes, whereas repressive H3K27 trimethylation is observed only at Polycomb‐repressed poised genes. H3K36 trimethylation is enriched at active genes but not at paused or poised genes, consistent with a role in productive elongation. PRC, Polycomb repressive complex; RNAPII, RNA polymerase II; TSS, transcription start site.
Low levels of α‐amanitin‐sensitive transcripts are produced from bivalent genes, and transcripts are polymerized and spliced to at least the first exon junction (Stock et al, 2007). This contrasts with the transcriptional activity of paused polymerases, which mainly synthesize short transcripts of a 10–50 nucleotides of base pairs (Rougvie & Lis, 1988). RNAPII at bivalent genes is characterized by an unusual set of post‐translational modifications. It contains high levels of Ser 5P but no detectable Ser 2P in spite of elongation. In addition, the binding of an antibody that preferentially recognizes non‐phosphorylated Ser 2 residues (8WG16) is unexpectedly low or even absent (Sidebar A
The absence of Ser 2P at poised RNAPII complexes is likely to disrupt co‐transcriptional RNA maturation and histone modification to help maintain the silent state, whereas binding and transcription by RNAPII counteract Polycomb‐mediated chromatin condensation (Fig 2B). Low levels of mRNA from bivalent genes can be detected in microarray analyses (Guenther et al, 2007; Mikkelsen et al, 2007), suggesting that transcript maturation occurs to a limited extent. H3K36 trimethylation—a histone modification that is associated with productive elongation—is not detected in coding regions of bivalent genes (E.B., H. Kimura and A.P., unpublished observations), which is consistent with the disruption of factor recruitment. Similarly, priming of inducible primary response genes in macrophages is characterized by RNAPII transcription through coding regions in the absence of RNA splicing (Hargreaves et al, 2009).
The mechanisms that prevent the phosphorylation of RNAPII Ser 2 at primed genes could involve uncharacterized CTD modifications or conformations that uncouple transcription from productive gene expression (Figs 1,2,3). A CTD modification might prevent Ser 2 phosphorylation by obscuring target sites, by affecting the structure of the CTD or by inducing the recruitment of specific factors. Conversely, the activity of CTD phosphatases that target specific genes—such as SCPs—could actively deplete Ser 2P. The modulation of the configuration of the CTD by isomerases such as Pin1 might regulate RNAPII poising, but their role at bivalent genes is yet to be explored.
Interplay between PRCs and poised RNAPII
In the prevailing model of PRC regulation, PRC2 is recruited to specific genomic locations and catalyses H3K27 trimethylation, thereby creating binding sites for PRC1, which then ubiquitinates H2AK119 (Cao et al, 2002). However, the distribution of the PRC1 ubiquitin ligase Ring1B and monoubiquitinated H2AK119 across individual bivalent loci in mouse ES cells unexpectedly parallels the occupancy of Ser 5P RNAPII more closely than that of PRC2 or H3K27 trimethylation (Stock et al, 2007). This suggests a direct role for PRC1 or H2AK119 monoubiquitination in restraining or recruiting poised RNAPII, or a dependency of Ring1B recruitment on RNAPII.
The removal of the catalytic PRC1 components Ring1A and Ring1B in ES cells causes the depletion of H2AK119 monoubiquitination and a derepression of bivalent genes ( Jorgensen et al, 2006; Stock et al, 2007). The CTD conformation of RNAPII at poised genes also changes in unexpected ways (Stock et al, 2007). Derepression is not accompanied by changes in Ser 5P levels, implying that the recruitment of RNAPII to promoters is not enhanced in the absence of PRC1. Ser 2P levels are not increased, suggesting that derepression occurs without the conversion of restrained, poised polymerases into typical, Ser 2P‐dependent elongation complexes. Ring1 depletion does, however, enhance 8WG16 binding to levels similar to those at active genes, raising the possibility of a mechanistic interaction between the conformation of the CTD and the recruitment of Ring1B or H2A monoubiquitination. By contrast, the depletion of 2A‐HUB—an H2A ubiquitin ligase that represses chemokine genes in a monocyte–macrophage cell line—correlates with increased Ser 2P at derepressed genes, indicating a release into Ser 2P‐dependent elongation (Zhou et al, 2008).
The correlation between Ring1B and RNAPII binding at PRC‐repressed genes in ES cells, and the change in RNAPII on Ring1B depletion (Stock et al, 2007), raise the intriguing possibility that PRC1 might directly modify RNAPII or crucial poising components to restrict RNAPII processivity at developmental regulator genes (Fig 2B). Interestingly, non‐consensus CTD repeats—which are present in mammals but not in yeast—have lysine residues, which are potential targets for ubiquitination or methylation, instead of the canonical Ser 7. We propose that the direct modification of RNAPII CTD lysines by PRCs could integrate Polycomb silencing with the establishment of a poised RNAPII configuration. PRC1 components have been shown to be absent from a subset of bivalent genes, adding yet another level of complexity to the regulation of genes that control developmental processes (Ku et al, 2008).
The modification of RNAPII confers robust control of gene expression levels by regulating the selective recruitment of chromatin and RNA processing factors. Preloading RNAPII at inactive genes primes them for subsequent activation without leading to immediate expression. RNAPII can be restrained at promoter‐proximal regions in the absence of modifications that are important for the onset of productive elongation. Alternatively, polymerases can transcribe through coding regions without the modifications that are necessary for the recruitment of chromatin modifiers or the RNA processing machinery. At developmental regulator genes in ES cells, PRCs control the poised state of RNAPII by influencing chromatin modification and RNAPII configuration. Antisense RNA transcription around promoters is widespread across the mouse ES cell genome (Seila et al, 2008) and represents an additional layer of gene expression regulation that is enabled by RNAPII. The identification of regulators of RNAPII states across the genome and of their mechanisms of action and functional interactions with chromatin modifiers, PRCs and RNA processors, will be important for understanding the mechanisms of gene regulation that control the responses to environmental stimuli and programme gene expression during cell commitment (Sidebar B).
Sidebar B | In need of answers
What are the enzymatic activities (kinases, phosphatases, other) that establish the different RNAPII states?
What are the key differences between RNAPII pausing at quickly inducible genes and RNAPII poising at PRC‐dependent developmental regulator genes?
What is the processing status of RNAs produced by paused and poised RNAPII complexes?
What are the kinetics of transcriptional pausing and poising?
What are the mechanistic relationships between PRC activity and RNAPII function?
Are there additional modifications of CTD residues, what is their function and how are they controlled? What are the roles of Ser 7 phosphorylation? Are non‐canonical CTD residues —which are found in higher eukaryotes— also modified?
We are grateful to J.K. Stock, C. Ferrai, I. de Santiago, M. Chotalia and A. Möller for critical reading of the manuscript, and to D. Eick and H. Kimura for sharing reagents and unpublished data. We thank the Medical Research Council (UK) for funding.
- H2A histone ubiquitin ligase
- carboxy‐terminal domain
- major histocompatibility complex, class II, DRα
- DRB‐sensitivity inducing factor
- embryonic stem
- transcription factor IIF‐associated CTD phosphatase
- histone methyltransferase
- negative elongation factor
- protein interacting with NIMA‐1
- Polycomb repressive complex
- positive transcription elongation factor b
- RNA polymerase II subunit B1
- RE‐1 silencing transcription factor
- really interesting new gene
- regulator of transcription 1
- small CTD phosphatase
- suppressor of variegation, enhancer of zeste and trithorax 1
- suppressor of sua7 protein 2
- TATA‐binding protein
- transcription factor IIH
- Copyright © 2009 European Molecular Biology Organization
Ana Pombo and Emily Brookes