A combinatorial view of old and new RNA polymerase II modifications

ABSTRACT

The production of mRNA is a dynamic process that is highly regulated by reversible post-translational modifications of the C-terminal domain (CTD) of RNA polymerase II. The CTD is a highly repetitive domain consisting mostly of the consensus heptad sequence Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7. Phosphorylation of serine residues within this repeat sequence is well studied, but modifications of all residues have been described. Here, we focus on integrating newly identified and lesser-studied CTD post-translational modifications into the existing framework. We also review the growing body of work demonstrating crosstalk between different CTD modifications and the functional consequences of such crosstalk on the dynamics of transcriptional regulation.

Introduction

Precisely tuned gene expression is crucial for maintenance of organismal homeostasis and for response to external stimuli. RNA Polymerase II (RNAPII) is a large multi-subunit enzymatic complex that is essential for constitutive expression of mRNA and small nuclear RNAs as well as their rapid production in response to altering cell signaling events. While RNAPII is highly conserved across eukaryotes, the C-terminal domain (CTD) of its largest catalytic subunit, Rpb1, has evolved to harbor unique features across species [1,2]. The CTD is highly repetitive, unstructured, and of low diversity [3]. The Rpb1 CTD consensus heptad sequence (Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7) is highly conserved as exemplified by the similarity of the Sacccharomyces cervisiae CTD to the first 26 proximal repeats of the human CTD. However, the CTD has varying divergence across eukaryotes with higher eukaryotes evolving a substantially longer CTD following the common ancestor of Metazoa [4]. The distal CTD in higher metazoans has maintained the heptad repeat structure, yet diverged from the consensus heptad sequence [1,5–8] (Figure 1). Recent work studying the ability of RNAPII to induce liquid phase separation suggest that organisms with increased CTD length harbor more non-consensus repeats to prevent protein aggregation of the CTD while still allowing phase separation creating transcriptional hubs [9–12].

Figure 1.

Schematic of the CTD. The RNAPII CTD of Saccharomyces cerevisiae, Drosophila melanogaster, and Homo sapiens are presented. Each heptad repeat is represented as a new line. Conserved repeats are in green, non-conserved amino acids are in bold, non-consensus repeats that can be uniquely modified are in yellow.

Although the CTD is a relatively small appendage of a massive molecular machine, it has a profound significance on the regulation of gene expression. The unstructured nature of the CTD hypothetically allows for increased protein-protein interactions with major transcriptional effector molecules, including the enzymes mediating or removing posttranslational modifications (PTMs; writers and erasers), and the proteins containing specific recognition domains for PTMs (readers). Many CTD-modifying or -recognizing proteins essential for regulating the process of transcription are highlighted in this review.

The most extensively studied modification of the CTD is the dynamic and reversible phosphorylation of serine residues, specifically of Ser2 and Ser5; the roles of phospho-Ser2 and -Ser5 in the regulation of transcription have been reviewed previously [5,7,13–16]. However, on-going research in the field has shown that every amino acid in the consensus heptad repeat, as well as arginine and lysine residues of non-consensus repeats, can be modified during transcription. Recognized CTD PTMs include phosphorylation of serine, tyrosine, and threonine, O-GlcNAcylation (O-GlcNAc), proline isomerization, arginine and lysine methylation, arginine citrullination, and lysine acetylation (Table 1). The repertoire of PTMs that have been attributed to the CTD is often referred to as a “CTD code”, in which certain modifications are attributed to particular functions during specific steps of the cycle (Figure 2).

Table 1.

CTD modifications, their functions and interacting partners.

Modification	Function	Reader	Writer	Eraser	Reference
Tyrosine 1 phosphorylation	Prevention of pre-mature termination, CTD stability, enhancer transcription		c-Abl		[79,80,108,163,166]
Threonine 4 phosphorylation	Transcription elongation, termination, processing of histone mRNA, chromatin remodeling		Plk1/3, CDK9		[167–170]
Proline Isomerization	cis-trans isomerization of CTD prolines, regulates activity of serine 5 phosphatases	Ssu72, FCP1	Pin1		[149–151,153,154,171]
Arginine citrullination	Promoter-proximal pause release	P-TEFb	PADI2		[109]
Arginine methylation	Transcription termination, terminal R-loop resolution, snRNA and snoRNA expression	SMN, TDRD3	PRMT5, PTRM4/CARM1		[27,28,172]
Lys Ubiquitination	Degradation of RPB1	rpfl/hNedd4	pVHL		[173,174]
O-GlcNAc	Assembly of PIC		OGT	OGA	[19–23,61]
Serine 2 phosphorylation	Promoter-proximal pause release, elongation, termination, splicing	PAF1, SPT6, TCERG1, U2AF65-Prp19, SET2, HDAC/HAT, SCAF8, SCAF4	CDK9(P-TEFb), CDK12, CDK13	Ssu72, FCP1	[30,37,42,103,122,152,157,175–190]
Serine 5 phosphorylation	mRNA capping, promoter-proximal pausing, chromatin remodeling, ncRNA transcription termination, splicing, prevention of premature transcription termination	DYRK1a, SCP1, SCP4, CDC14, MLL1/2, Guanyltransferase, Pin1	CDK7, CDK8, CDK9, CDK12, CDK13	Ssu72, RPAP2	[74,76,151,153,155,191–198]
Serine 7 phosphorylation	snRNA expression, Interaction with Integrator, promoter-proximal pausing		CDK7, CDK9	Ssu72	[152,199–202]

Figure 2.

The cycle of CTD modifications. Key CTD post translational modifications of each major phase of transcription are shown, with the average duration of each phase noted. The rate of transcriptional initiation is not listed as a time range as rates of initiation vary dramatically depending on if the mediator complex is already resident at the gene promoter. Timing for phases is from eukaryotes in general [31,32,38–44,46-50]. Circles denote phosphorylation and diamonds indicate O-GlcNAc unless otherwise noted.

Although many modifications are highly conserved, their time of occurrence and associated functions can vary across species. For example, the role of Tyr1 and Ser7 phosphorylation vary considerably between mammals and yeast. In mammals, both modifications control early steps of transcription, but in yeast, Ser7 phosphorylation acts both early in transcription and as a termination signal and Tyr1 phosphorylation has primarily been linked to control of transcription termination [5,17,18]. Other modifications have only been described in mammals, including O-GlcNAcylation [19–23], lysine acetylation [4,24–26] and arginine methylation [27,28], and the extended PTM repertoire suggests a need for enhanced regulation in more complex organisms.

Increasing evidence indicates that individual PTMs can communicate and act to modulate each other’s presence and function. This crosstalk expands the regulatory capacity and complexity of the CTD and allows RNAPII to rapidly and transiently react to external stimuli. Both positive and negative crosstalk mechanisms have been described, either promoting or prohibiting the occurrence of modifications, or enhancing or impairing reading of the modifications’ “code”. In this review, we focus on PTMs of the mammalian CTD during mRNA synthesis, including those that are less studied, and discuss the emerging significance of PTM crosstalk within the CTD.

Timing of individual CTD modifications

Transcription of genes coding for mRNA takes place in four key phases: initiation, promoter-proximal pausing, elongation, and termination [29].The dynamic, reversible PTMs of the CTD mark the various stages of transcription (Figure 3), from an uninitiated polymerase to transcription termination [30]. This is key to allowing transcription to adapt to specific needs, for example transcription must progress at fast rates in response to external stimuli as well as during the rapid cell divisions of early development [31,32]. Although CTD modifications play a key role in the regulation of transcription rates, it is important to note that while not discussed in this review other features of genes, such as nucleosome position, DNA sequence, DNA structure, and co-transcriptional processes, are also involved [33–37]. Studies on bulk transcription rates in human cells indicate that the average gene can be transcribed on the scale of several minutes [38–44]. However, this varies dramatically with gene size; indeed, the largest gene in the human genome takes 16 hours to transcribe [45]. Development of new single molecule techniques for studying individual transcripts may allow further elucidation of the rate of individual phases of transcription and the role of PTMs in regulating these phases [46–51].

Figure 3.

Key steps of transcription. A and B) An unmodified RNAPII is recruited to the transcription start site (TSS) by the Mediator complex. O-GlcNAcylation of Ser5 and 7 is performed by OGT. C) The pre-initiation complex (PIC) is formed and its CDK7 subunit phosphorylates Ser5 and 7 after O-GlcNAc is removed by OGA. D) Ser5 and 7 phosphorylation releases the polymerase from interactions with Mediator and the PIC. Following release, the polymerase pauses proximally to the promoter, marked by Tyr1 phosphorylation by c-Abl, acetylation of Lys7 by p300, and citrullination of non-consensus Arg1810 by PADI2. Pausing is facilitated by the recruitment of NELF and DSIF. E) The polymerase is released from pausing by phosphorylation of Ser2 by the CDK9 subunit of PTEF-b, as well as the phosphorylation of DSIF and removal of NELF. F) Once productive elongation begins, CDK12 and CDK13 maintain the phosphorylation of Ser2, and phosphorylation of Ser5 is removed by the Ssu72 phosphatase. G) As the polymerase reaches the 3ʹ end of the gene, Thr4 phosphorylation by PI3 K occurs, as well as methylation of Arg1810 by PRMT5. H) When the polymerase is in the proximity of a poly-adenylation site (PAS), it interacts with the cleavage and polyadenylation complex (CPAC) via Ser2 phosphorylation, allowing for cleavage of the mRNA from the polymerase. I) The polymerase is then removed from the DNA and the remaining Ser2 phosphorylation is removed by FCP1. J) A unmodified polymerase is free to reinitiate new rounds of transcription.

Transcriptional initiation

Initiation is the earliest stage of the transcription cycle and involves several key steps: initial recruitment of the polymerase to a promoter, proper formation of the transcription preinitiation complex (PIC), and promoter escape. Classically, transcriptional initiation was believed to be associated with an unmodified CTD [35]. However, an increasing body of work has shown that less-studied PTMs are critical to steps in early transcriptional initiation [17,20,26,52,53].

First, a polymerase with an unmodified CTD interacts with Mediator, the transcriptional co-activator complex [54,55]. This is thought to function as a bridge between general transcription factors and RNAPII, and facilitate the formation of the PIC [56–59]. Proper formation of the PIC requires specific CTD PTMs, including O-GlcNAcylation of Ser5 and Ser7 subsequent to polymerase binding to Mediator [20,23,52]. O-GlcNAcylation is a highly transient event mediated by the O-GlcNAc transferase (OGT) and rapidly reversed by the N-acetylglucos-aminidase (OGA), a member of the PIC [19]. The removal of O-GlcNAc after PIC formation is required for phosphorylation of Ser5 and Ser7, which mark the transition to elongation, supporting the model that O-GlcNAc of RNAPII is transient and likely only exists for the formation of the PIC [19,22]. Notably, O-GlcNAc modifications have only been described in mammals and are likely unique to vertebrates, as the O-GlcNAc enzymes are not present in lower eukaryotes, including yeast [60,61].

To begin early elongation, RNAPII must escape interactions with the Mediator complex and the PIC. Notably, many polymerases will fail to escape the promoter and will be turned over through abortive termination [35,62]. The interaction among RNAPII, the PIC, and Mediator are disrupted by phosphorylation of Ser5 and Ser7 by TFIIH, a member of the PIC, via its CDK7 subunit, which frees RNAPII to initiate the early steps of elongation [63–68]. Interestingly, the kinase activity of TFIIH is not required for initiation in vitro, although it is required for early elongation [69–73]. Ser5 phosphorylation is also critical for recruitment of enzymes involved in capping the nascent RNA, which protects it from degradation [74–76]. This function is supported by the recruitment of Dichloro-1-B-D-ribofuranosyl-benzimidazole Sensitivity-Inducing Factor (DSIF), which assists in RNA capping and recruitment of the Negative Elongation Factor (NELF) [42,57].

Two other less studied PTMs play a role during transcriptional initiation: phosphorylation of Tyr1, one of the most highly conserved CTD residues, and methylation of Lys7 [17,26,53,77–80]. Phosphorylation of Tyr1 has been associated with RNAPII occupying enhancers and promoters, and more recently attributed to driving specificity of CDK kinase activity on CTD repeats (discussed below). Tyr1-phosphorylated RNAPII is also involved in the production of enhancer RNAs that occur specifically in the anti-sense direction of gene bodies [17,77–80]. The mechanism and role of lysine methylation remain largely unknown, but initial work indicates that it is involved in preventing Lys7 acetylation, and may negatively regulate transcription [26,53]. Of note, Lys7 exists on only eight non-consensus repeats in the distal metazoan CTD, thus this lysine methylation is a unique feature of higher eukaryotic CTDs.

Promoter-proximal pausing

After promoter escape, RNAPII can pause proximal to the promoter, at an average of 25–50 bp downstream of the transcriptional start site (TSS) [57,81–85]. This phenomenon was initially believed to occur at a small number of genes or viruses such as heat shock genes in Drosophila [86–88] and the integrated HIV provirus [81]. It is now recognized that promoter-proximal pausing is a feature of RNAPII transcription of most genes, as disruption of the release from pausing abrogates transcription of nearly all genes [42,89–92]. Not all paused polymerases will continue into productive elongation, and some paused polymerases will be removed and replaced [42,93–96]. The turnover of paused polymerases may play a critical role in regulating the rate of transcription. Whether CTD PTMs play a role in RNAPII turnover remains to be determined.

Promoter-proximal pausing is established and maintained by the interaction of the CTD with negative elongation factors NELF and DSIF [92,97–100]. Paused polymerases are classically characterized by high levels of Ser5 and Ser7 phosphorylation, but newer research has highlighted the role of additional PTMs. The transition of the polymerase from a paused state to an actively elongating complex, called pause release, is catalyzed by the recruitment of the positive transcriptional elongation factor b (P-TEFb) [37,57,89,90,101–105]. P-TEFb is composed of a regulatory cyclin and catalytic CDK9 component, which phosphorylates Ser2 residues to enhance polymerase processivity [102–104,106]. The ability of P-TEFb to phosphorylate Ser2 is primed by previous phosphorylation events at Ser7 [107] and Tyr1 [108]. A recent study in a breast cancer cell line identified citrullination of Arg1810 by peptidyl arginine deiminase 2 (PADI2), a calcium-dependent enzyme, as a key CTD modification for recruitment of P-TEFb to the paused polymerase of genes involved in cell proliferation [109]. It is important to note the breast cancer cell line studied expressed high levels of PADI2 and the role of citrullination in cells with normal levels of PADI2 expression remains to be fully explored. Given that PADI2 is a calcium dependent enzyme [109], it is intriguing to hypothesize that citrullination may act in a gene-selective manner modulating promoter-proximal release in response to calcium-induced signaling cascades. To date, there is no known mechanism for reversal of arginine citrullination, leading to questions about what happens to citrullinated RNAPII following pause release [109–111]. P-TEFb also phosphorylates DSIF and NELF; phosphorylation of NELF leads to its dissociation from the polymerase complex, whereas phosphorylation of DSIF turns DSIF into a positive elongation factor [100–103,112-115].

Paused polymerases are also highly acetylated at Lys7 residues, mediated by the acetyltransferase p300 [24,25]. Lys7 acetylation serves to recruit the RPRD complex, which includes RPAP2, which serves as a Ser5 phosphatase, and HDAC1, which deacetylates Lys7 [24]. Phosphorylation of Ser5 has been shown to inhibit the ability of P-TEFb to phosphorylate Ser2 on the same heptad in vitro [107], so recruitment of RPRD may prime RNAPII for processive elongation. Furthermore, because Lys7 acetylation results in recruitment of its own deacetylase, Lys7 acetylation is limited to a very narrow window downstream of the TSS [24,25]. The fact that methylation of Lys7 may further restrict Lys7 acetylation underscores the notion that Lys7 acetylation acts at a very defined time during the transcription cycle [26,53].

Transcriptional elongation

Transcriptional elongation is characterized by a gradual loss of Ser5 and Ser7 phosphorylation and Lys7 acetylation [25,116]. In addition, there is an increase in Ser2 phosphorylation, beginning just downstream of promoter pausing and increasing gradually along the gene body, reaching maximal levels at the 3ʹ end of genes [117]. The rate of transcriptional elongation also increases along the body of the gene [32,42–44], and the accumulation of Ser2 phosphorylation along the gene appears to be regulated by the rate of elongation, as RNAPII mutants with slower elongation rates lead to increased Ser2 phosphorylation toward the 5ʹ end of genes [118].

Critical to pause release and transcriptional elongation is the recruitment of complexes mediating co-transcriptional processes, such as splicing or polyadenylation [31,119,120]. Interactions between the CTD of actively transcribing RNAPII and the spliceosome were thought to be mediated primarily through interactions with phosphorylated Ser5 [121]. However, the crystal structure of transcriptional elongation regulator 1 (TCERG1), which mediates interactions between RNAPII and the spliceosome, indicates that hyperphosphorylation of all three serine residues, Ser2, Ser5 and Ser7, are required for optimal interactions between RNAPII and the spliceosome during elongation [122–124]. Additionally, premature termination of transcription is inhibited by the interaction between the CTD phosphorylated on Ser2 and Ser5 and the human anti-termination proteins SCAF4 and SCAF8, members of the arginine/serine-rich splicing factor family [125]. Indeed, loss of Ser2 phosphorylation leads to increases in the usage of early alternative polyadenylation sites by RNAPII [126].

While the initial Ser2 phosphorylation is placed by P-TEFb/CDK9, maintenance of Ser2 phosphorylation throughout elongation is carried out by different kinases, specifically CDK12 and CDK13 [127–130]. Experiments knocking down CDK12 and CDK13 indicate that these kinases have individual as well as overlapping function, however, it is unlikely that they cooperate on individual genes [128,130–132]. How these kinases act to maintain Ser2 phosphorylation in a gene-specific manner is an area of ongoing research.

Transcriptional termination

Recent work has dramatically increased our understanding of the mechanism of transcriptional termination in mammalian systems. However, the exact role that CTD modifications play in termination in mammals remains elusive. There are two dominant models of mammalian termination; Allosteric, where conformational changes to RNAPII allow for termination, and the Torpedo model, where RNA polymerase continues transcribing until it is removed by XRN2 [133]. Which of these models is correct remains unclear, although recent work suggests that aspects of both occur together for efficient termination [134]. Termination occurs in two steps: cleavage/polyadenylation of mRNAs, and removal of RNAPII from DNA [133,135]. These steps occur separately, which suggests that identification of the correct polyadenylation site is necessary for efficient termination. Transcriptional termination is associated with a second pausing of the polymerase, this time at the 3ʹ end of the gene [136–140]. Whether 3ʹ pausing is required for termination remains controversial, and the cause remains elusive, although it may involve interactions of RNAPII with the cleavage and polyadenylation (CPA) complex and the polyadenylation signal [36,138,141].

Interactions between the CPA complex and RNAPII are mediated through the termination factor PCF11, which shows selective interaction with Ser2-phosphorylated RNAPII [142,143]. Additionally, Arg1810 di-methylated by the methyltransferase PRMT5 is recognized by SMN, a protein involved in spliceosome assembly and ribonucleosome biogenesis [27,144]. SMN in turn recruits Sentataxin (SETX), a DNA-RNA helicase required for cleavage of the mRNA from RNAPII [27]. Although the PRMT proteins needed for arginine methylation are conserved across eukaryotes, non-consensus repeats harboring arginine only exist in vertebrates [27,145]. Following cleavage of the mRNA, the exonuclease XRN2 removes RNAPII from chromatin [36,141]. Once the RNAPII is removed from the DNA, it is maintained in a hypo-phosphorylated state by the phosphatase FCP1, which allows efficient recycling of RNAPII into new rounds of transcription [146].

Crosstalk between individual CTD modifications

While CTD modifications have long been considered individually in a systematic “ON” and “OFF” exchange throughout the transcription cycle, recent examples of cooperativity between PTMs have uncovered new functions of individual PTMs and have increased the complexity of our model of transcriptional regulation (Figure 4). Analogous examples of PTM crosstalk, such as among histone modifications, demonstrate that the complex combinatorial nature of modifications can drive specificity and selectivity of interactions with reader proteins [147]. The CTD, similar to the histone tails, is structurally flexible and can support interactions between different PTMs within the same repeat or between different repeats. The CTD is, however, unique in its highly repetitive sequence, and it remains unclear how many repeats within the CTD carry actual modifications. Recent mass spectrometry studies to map phosphorylation of the CTD have started to address these issues. Interestingly, these studies show that the CTD is not heavily phosphorylated, with phosphorylation of Ser2 and Ser5 being predominant and that heptads with multiple phosphorylation events are rare [106,148]. One limitation of these heptad-specific studies is that only phosphorylation marks of the CTD have been mapped and do not address less studied CTD modifications. These studies are also performed on bulk RNAPII and modifications that may be very transient or unstable during sample processing may be difficult to detect and accurately quantify.

Figure 4.

Cross-talk in the CTD. A) Proline isomerization influences the ability of CTD phosphatases to remove Ser5 phosphorylation. In the trans configuration, SCP1 is able to de-phosphorylate Ser5. While in the cis configuration mediated by the proline isomerase Pin1, SSU72 is able to de-phosphorylate Ser5. B) Acetyled non-consensus lysine residues interact with the reader proteins RPRD1A (1A) and RPRD1B (1B), which recruit the Ser5 phosphatase RPAP2. C) Phosphorylation of Tyr1 by c-Abl directs the kinase activity of the CDK9 subunit of PTEF-b to Ser2 by preventing it from phosphorylating Ser5 of nearby repeats. D) Premature termination is prevented by the recognition of Ser2 and Ser5 phosphorylation by SCAF4 and SCAF8, which block recruitment of CPAC to alternative polyadenylation sites (PAS).

Proline isomerization couples CTD phosphorylation to RNA processing events

A classical example of CTD crosstalk is the role of proline isomerization in directing serine/threonine phosphorylation. Prolines are highly conserved at position 3 and 6 of consensus repeats, and are also widely found in non-consensus repeats (Figure 1). Although the dominant state of the proline within the CTD is the more energetically stable trans-configuration, the peptidyl-prolyl isomerase Pin1 is able to recognize phosphorylated serine-proline motifs (Ser2-Pro3 or Ser5-Pro6) and isomerize them from the trans- to the cis-configuration [149–151].

In turn, the conformation of prolines can affect CTD phosphatases and influence the phosphorylation status of Ser2 and Ser5. Indeed, Pin1 peptidyl-prolyl isomerase activity influences the action of the human serine phosphatases Ssu72 and SCP1 [152–156] (Figure 4a). Molecular modeling studies predict that the presence of cis-proline significantly disrupts the active site of SCP1, thus making SCP1 preferentially act on serine residues proximal to trans-prolines [154]. Ssu72 is ubiquitously expressed and acts only on CTD repeats when prolines are in cis conformation [153,155,157].

While effects on serine phosphorylation are clear, it is unclear yet when isomerization occurs in the transcription cycle. Notably, Ssu72 functions as a Ser5 phosphatase only after promoter-proximal pause release [158–160], suggesting Pin1 functions during transcriptional elongation. Interestingly, overexpression of Pin1 leads to overall increased levels of Ser2 and Ser5 phosphorylation and a global shut-down of transcription [150]. These results support the hypothesis that the timing of proline isomerization plays a key role in the progression of transcription, but further studies are required to fully elucidate the function of proline isomerization in gene expression.

Reading of lysine acetylation leads to CTD de-phosphorylation

Higher eukaryotes have evolved a longer CTD and incorporated a varying number of non-consensus repeats carrying Lys7 residues [4]. Lys7 can be methylated and acetylated at different times during the transcription cycle [24–26,53]. Lys7 acetylation is functionally required for activation of signal response genes, e.g. in response to epidermal growth factor signaling, an essential function in the development of multicellular organisms [25].

A recent quantitative mass spectrometry analysis connected Lys7 acetylation with the preferred recruitment of the RPRD1 complex [24]. RPRD1A and RPRD1B and their yeast homologs were previously shown to interact with the Ser2-phosphorylated CTD, with increased affinity for repeats that are dually phosphorylated, and play a role in transcriptional termination [161,162]. However, in the more recent study Lys7 acetylation was shown to enhance RPRD1A and B interaction by providing new electrostatic interactions with the CTD [24] (Figure 4b).

As the RPRD1 protein complex also contains RPAP2, a serine 5 phosphatase, experiments using deacetylase inhibitors were performed to test the effect of increased Lys7 acetylation on Ser5 phosphorylation. Indeed, increased lysine acetylation led to decreased Ser5 phosphorylation, consistent with enhanced recruitment of RPAP2 to the acetylated CTD [24]. Interestingly, knockdown of RPRD1B also increased Lys7 acetylation, and indeed, the class I deacetylase HDAC1 was found associated with the RPRD1 complex [24]. This establishes not only a new link between Lys7 acetylation and Ser5 dephosphorylation, it also shows that certain modifications can autoregulate by recruiting an eraser protein as part of a reader complex. Notably, Lys7 acetylation occurs exclusively in distal repeats, while Ser5 phosphorylation is found across the CTD [26]. The crosstalk between Lys7 acetylation and Ser5 phosphorylation likely occurs between different repeats, in accordance with biochemical studies which showed that Ser5 phosphorylated repeats cannot be acetylated, and vice versa [24,25].

Tyrosine phosphorylation is required for RNAPII elongation via Ser2 phosphorylation

While Lys7 acetylation negatively regulates Ser5 phosphorylation, Tyr1 phosphorylation positively influences Ser2 phosphorylation. Tyr1 phosphorylation has been implicated in regulating transcriptional termination in yeast, specifically by preventing binding of the termination factor Nrd1 [18]. However, in mammals, Tyr phosphorylation occurs primarily at the 5ʹ end of genes [80,163]. A recent study showed that Tyr1 phosphorylation by c-Abl selectively directs P-TEFb/CDK9 to phosphorylate Ser2 [108] (Figure 4c). In biochemical studies, CDK9 can phosphorylate both Ser2 and Ser5 [107], but in vivo it specifically targets Ser2 [105]. Notably, the in vitro assessment of CDK9 function was performed using short CTD peptides, and thus may not account for spatially separated interactions between different repeats.

Prevention of early termination is mediated by dual-phosphorylation of Ser2 and Ser5

Transcriptional termination depends on precise polyadenylation of mRNAs, which is complicated by the existence of multiple early polyadenylation sequences in human genes [164,165]. Thus, extraneous polyadenylation sites must be ignored by the actively transcribing polymerase complex. Recent work showed that the proteins SCAF4 and SCAF8 prevent premature termination by inhibiting recognition of early alternative polyadenylation sites [125] (Figure 4d). In vivo, SCAF4 and SCAF8 co-immunoprecipitate with actively elongating RNAPII that is hyperphosphorylated at Ser2/5/7 and at Tyr1 and Thr4. In vitro experiments with purified CTD heptads showed that SCAF4 and SCAF8 had strong preference for heptads that are dually-phosphorylated at Ser2 and Ser5. This suggests that by recognizing dually-phosphorylated heptads, SCAF4 and SCAF8 identify polymerases early in the process of elongation, preventing premature recruitment of the CPA complex. This is supported by recent work on the function of CDK12, a kinase that supports Ser2 phosphorylation during elongation. Loss of CDK12, and thus loss of Ser2 phosphorylation, led to use of early polyadenylation sites [126].

Conclusions

CTD modifications are key regulators of each step of eukaryotic transcription, and our emerging understanding of the interplay between CTD modifications sheds light on how transcription can be rapidly and dynamically responsive to stimuli. Deciphering the interactions between modifications will lead to an improved understanding of their timing and distribution during the transcriptional cycle. It will also clarify the dynamics of reader protein recruitment, and could lead to new approaches to therapeutically interfere with these dynamics. New research is also beginning to reveal the role of modifications of the CTD at non-consensus sites, which underlie the enhanced transcriptional complexity in higher eukaryotes.

However, the field is hampered by technical challenges, such as antibodies with limited specificity, and the inability of mass spectrometry to provide resolution at the level of a single repeat. In addition, while recent advances such as cryo-EM have greatly improved the ability to capture structural information about RNAPII, the CTD structure in the various phases of transcription remains elusive because of its flexible nature. Future studies directly addressing these issues are necessary to fully capture the complexity of transcription regulation and the important role of CTD modifications in this process.

Funding Statement

This work was supported by the National Institute of Allergy and Infectious Diseases [1R61AI140465-01]; National Institute of Allergy and Infectious Diseases [2R37AI083139-10]; National Institute of Allergy and Infectious Diseases [1R61DA048444-01]; National Institute on Drug Abuse [5R01AI097552-03]; National Institute on Drug Abuse [1R01DA043142-01]; National Institutes of Health [5DP1DA038043-03]; Tobacco-Related Disease Research Program (US) [T30DT1006]; National Institute of Diabetes and Digestive and Kidney Diseases (US) [1R01DK123746-01].

Disclosure statement

No potential conflict of interest was reported by the authors.