The point of no return: The poly(A)-associated elongation checkpoint

abstract

Cyclin-dependent kinases play critical roles in transcription by RNA polymerase II (pol II) and processing of the transcripts. For example, CDK9 regulates transcription of protein-coding genes, splicing, and 3′ end formation of the transcripts. Accordingly, CDK9 inhibitors have a drastic effect on the production of mRNA in human cells. Recent analyses indicate that CDK9 regulates transcription at the early-elongation checkpoint of the vast majority of pol II-transcribed genes. Our recent discovery of an additional CDK9-regulated elongation checkpoint close to poly(A) sites adds a new layer to the control of transcription by this critical cellular kinase. This novel poly(A)-associated checkpoint has the potential to powerfully regulate gene expression just before a functional polyadenylated mRNA is produced: the point of no return. However, many questions remain to be answered before the role of this checkpoint becomes clear. Here we speculate on the possible biological significance of this novel mechanism of gene regulation and the players that may be involved.

Abbreviations

CDK

cyclin-dependent kinase

DSIF

DRB sensitivity-inducing factor

EEC

early-elongation checkpoint

EC

elongation complex

GRO-Seq

global run-on sequencing

NELF

negative elongation factor

NET-Seq

nascent elongation transcript sequencing

pol II

RNA polymerase II

pol II CTD

RNA polymerase II carboxyl terminal domain

PRO-Seq

precise nuclear run-on sequencing

TC

termination complex

TSS

transcription start site.

Pol II Traffic Management

Pol II is paused at promoter-proximal, early-elongation checkpoints (EECs) by DRB-sensitivity-inducing factor (DSIF) and negative elongation factor (NELF) (Fig. 1A).¹ Pausing is overcome by P-TEFb, comprising CDK9 and a Cyclin T¹, which phosphorylates the Spt5 subunit of DSIF and the E subunit of NELF. This causes release of NELF and converts DSIF into a positive elongation factor.^1-3 P-TEFb also phosphorylates Ser2 of the Tyr1/Ser2/Pro3/Thre4/Ser5/Pro6/Ser7 repeats in the mammalian pol II carboxyl-terminal domain (CTD), which allows the CTD to mediate recruitment of RNA processing and elongation factors.^3-5 This checkpoint could help check that key elongation and RNA processing factors are recruited before pol II goes ahead, in analogy to checking you have your keys and your wallet before you shut the front door behind you in the morning. It also provides a mechanism to control gene expression through regulating release of pol II into productive elongation, eg in response to activation. For example, c-myc, NF-kB, and HIV Tat recruit P-TEFb to activate transcription elongation,¹ while the Herpes simplex-1 (HSV-1) ICP22 protein inhibits CDK9 activity to downregulate host cell gene expression.^6-8 In addition, pausing at this checkpoint can ensure synchronous activation during development⁹ and control of signaling networks in embryonic stem cells.¹⁰

Figure 1.

Pol II traffic management. (A) The recruitment of the RNA polymerase II (pol II) by general transcription factors (GTFs) at the promoter transcription start site (TSS) is followed by the phosphorylation of Ser5 and Ser7 of the pol II CTD by TFIIH. The CTD domain is also phosphorylated at Tyr1 at this point. Recruitment of the negative elongation factor (NELF) and DRB-sensitivity-inducing factor (DSIF) halts pol II soon after initiation at the early elongation checkpoint (EEC). Release from the EEC is controlled by P-TEFb, which phosphorylates Ser2 of the pol II CTD, the E subunit of NELF, and the Spt5 subunit of DSIF. Phosphorylation of NELF results in its removal from the pol II whereas phospho-DSIF becomes a positive elongation factor. These two events allow the recruitment of elongation factors and RNA processing factors (_EC) and promote elongation. In the presence of CDK9 inhibitors, the pol II cannot negotiate this checkpoint anymore and therefore cannot proceed. Here and in subsequent figures: Orange line: mRNA (orange filled circle: mRNA cap, orange filled square: poly(A) tail); orange dashed line: degraded mRNA; light green filled circle: phosphorylation; pink filled circle: Tyr1 phosphorylation; yellow filled circle: Ser2 phosphorylation; brown filled circle: Thr4 phosphorylation; red filled circle: Ser5 phosphorylation; blue filled circle: Ser7 phosphorylation. (B) In normal conditions, the elongating pol II transcribes up to a poly(A)-associated checkpoint where P-TEFb (and possibly other kinase including CDK12) is required for phosphorylating one or more proteins. The negotiation of this checkpoint allows the pol II to transcribe beyond the poly(A) site, after which the elongation complex (_EC) is exchanged for the termination complex (TC), and cleavage and polyadenylation can occur. Phosphorylation of Tyr1 and Thr4 (Tyr1P and Thr4P) of the pol II CTD increases when the pol II approaches the poly(A) site and an increase in Tyr1P, Ser2P, and Thr4P is observed in the terminating pol II. Ser5P is slightly reduced after the poly(A) site and the Ser7P remains unchanged. (C) In presence of CDK9 inhibitors, the pol II that already negotiated the early-elongation checkpoint before the drugs entered the cell can still elongate until it reaches the poly(A)-associated checkpoint. In the absence of the kinase(s) activity pol II terminates prematurely, close to the poly(A) site. The fate of the mRNA, eg its cleavage and polyadenylation, is unknown. In the presence of CDK9 inhibitors, the phosphorylation pattern of the CTD of pol II upstream of the poly(A) site is similar to the pattern downstream of the poly(A) site in normal conditions, except that Ser2P does not increase and Ser7P is slightly increased.

Global Run-On Sequencing (GRO-seq) analysis¹¹ of actively-engaged pol II after treatment of HeLa cells with the CDK9 inhibitors DRB, KM05283, and Flavopiridol indicates that the vast majority of pol II-transcribed genes have a kinase-dependent early-elongation checkpoint.^12,13 The CDK9 inhibitors also induce an increase in pol II levels just downstream of the transcription start site (TSS), which is thought to result from pol II piling up because it is unable to negotiate its way through the early elongation checkpoint.^14,15 However, there are also indications that pol II can terminate soon after initiation¹⁶ and inhibition of termination directly or indirectly by the inhibitors used could cause a build up of pol II near the TSS.

Pol II that has passed the early-elongation checkpoint before CDK9 inhibitors take effect is able to continue transcription,^12,13 indicating that sustained CDK9 activity is not required for pol II to elongate. This continued elongation allows the effect of the inhibitors on transcription at the end of long genes to be uncovered when drug treatment is short enough.¹² Surprisingly, we found that all CDK9 inhibitors we tested (KM05283, DRB, and Flavopiridol) cause abrupt termination of transcription close to the annotated terminal poly(A) sites when used at high concentration (100μM, 200μM, and 1μM, respectively). These results support the notion that a kinase-dependent transcription elongation checkpoint is associated with poly(A) sites in human cells. Inhibition of ectopically-expressed analog-sensitive CDK9 also affects transcription at the beginning and end of the KPNB1 gene,¹² implicating CDK9 in regulating checkpoints at both ends of protein-coding genes.

Outstanding questions

Are poly(A)-associated checkpoints universal in eukaryotes?

Our GRO-Seq analysis at the 3′ end of genes was limited to genes longer than 30 kb¹² and we cannot exclude that the poly(A)-associated checkpoint is a specific feature of long human genes. An important goal of future research will therefore be to determine whether transcription checkpoints linked to RNA 3′ end formation signals are a universal feature of actively-transcribed, pol II-dependent genes in human cells. It will also be important to establish whether the poly(A)-associated checkpoint is conserved in eukaryotes. During transcription of yeast genes, transcription elongation factors are exchanged for polyadenylation and termination factors when the poly(A) site is reached,^17-19 indicating that this RNA processing signal directs transitions at the end of the yeast pol II transcription cycle.^18-21 However, premature termination near the poly(A) site is not caused by depletion or inhibition of either of the 2 transcription elongation-associated yeast kinases, Ctk1 and Bur1.^17,20,22,23 In mammalian cells, pol II pauses and pol II CTD phosphorylation is altered after the poly(A) site,^12,24-27 indicating that a transition also occurs at this point in the eukaryotic transcription cycle. It is therefore possible that additional kinase-dependent steps in the transition(s) from an elongating to a terminating pol II have arisen to create a powerful poly(A)-associated checkpoint regulating expression of mammalian genes.

Which kinases and kinase targets regulate the poly(A)-associated checkpoint?

As KM05283, DRB, and Flavopiridol can inhibit kinases other than CDK9,^28-30 it is possible that more than one kinase is involved in poly(A)-associated checkpoint control. In addition, a relatively low level of Flavopiridol (0.2μM) specifically inhibits transcription downstream of the early elongation checkpoint without affecting transcription downstream of the poly(A) site,¹³ suggesting either that a different set of kinases/targets are operating at each checkpoint or that transcription through the EEC requires a higher level of target phosphorylation. This is further supported by the observation that specific inhibition of the kinase activity of ectopically expressed analog-sensitive CDK9 does not recapitulate fully the effect caused by CDK9 inhibitors.¹² CDK12, a pol II CTD kinase implicated in 3′ end formation of mRNAs³¹ is a likely additional candidate kinase regulating the poly(A)-associated checkpoint as it is inhibited in vitro by DRB and Flavopidirol²⁹ perhaps unsurprisingly given that the active sites of CDK9 and CDK12 are strikingly similar. However, additional/alternative kinases cannot be ruled out.

Termination of transcription at poly(A)-associated checkpoints likely involves premature loss of elongation factor(s) and/or gain of termination factors. Several critical factors are lost from the poly(A) region after drug treatment including Spt5, which is an elongation factor and CDK9 target, CstF64, which binds to the poly(A) signal in the RNA, Ssu72, which is a part of the polyadenylation complex and a CTD Ser5P phosphatase, and CDK9 itself.¹² The loss of these factors could be a cause or consequence of premature termination near the poly(A) site but indicates that association of a range of factors with this region of genes is highly dynamic. The loss of poly(A) factors suggests that cleavage/polyadenylation is no longer occurring. However, it is also possible that cleavage/polyadenylation is simply happening much more rapidly than usual. In this case, kinase activity would delay these processes, eg by regulating cleavage by CPSF73, possibly to provide a window to regulate mRNA 3′ end formation.

Dynamic modification of the pol II CTD orchestrates the recruitment of elongation-associated and RNA processing factors during transcription and is implicated in transcription termination.³ It is therefore possible that premature termination caused by the drugs is due to changes in CTD phosphorylation causing premature loss of elongation factors or premature gain of termination factors. Phosphorylation of Ser2 (Ser2P) upstream of the poly(A) site is not affected by CDK9 inhibitors, while phosphorylation of Tyr1 (Tyr1P), Thr4 (Thr4P), and Ser7 (Ser7P) increase and phosphorylation of Ser5 (Ser5P) somewhat decreases¹² (Fig. 1B-C). The CTD phosphorylation profile upstream of the poly(A) site after CDK9 inhibition largely mirrors the profile associated with termination downstream of the poly(A) site in untreated cells. These results suggest common mechanisms of termination affected by or affecting CTD modification in the presence and absence of CDK9 inhibitors. Consistent with the maintenance of CTD Ser2 phosphorylation, the poly(A)/termination factor, Pcf11, which binds Ser2P^3,5,32 remains associated with genes after CDK9 inhibition.¹² However, Pcf11 is not required for premature termination caused by CDK9 inhibitors, although it may be part of the machinery involved.

As transcriptionally-engaged pol II elongates in the presence of CDK9 inhibitors,^12,13 continued phosphorylation of CDK9 targets like Spt5 is not necessary or phosphorylation persists after the early-elongation checkpoint. Why then is kinase activity needed for elongation beyond the poly(A)-associated checkpoint? Three scenarios could account for this: 1) a phosphatase that targets phosphorylated elongation factors (eg Spt5P) may be active just before the checkpoint, necessitating de novo phosphorylation for elongation to continue, 2) phosphorylation of elongation factors recruited de novo may be necessary to transcribe past the poly(A) site, 3) phosphorylation abrogates the activity of termination factors recruited at the poly(A) site to delay termination until eg poly(A) selection is complete.

CDK9 inhibitors may also cause premature termination due to indirect effects on histone occupancy or modification of histones at the poly(A) region. For example, trimethylation of histone H3 lysine 36 (H3K36^m3) is linked to CTD Ser2 phosphorylation and CDK9 inhibition greatly reduces this histone modification.^3,33

We favor the hypothesis that CDK9-dependent phosphorylation of Spt5 plays a key role in negotiating poly(A)-associated checkpoints.¹² However, it seems likely that regulation of pol II elongation at these checkpoints involves more than one kinase and a range of kinase targets (Table 1).

Table 1.

Factors implicated in poly(A)-associated checkpoint control.

Putative candidates	Functions	References
Spt5	Elongation factor, subunit of the DSIF complex. Phosphorylated by CDK9.	12
Pol II CTD	Can be phosphorylated by CDK9 and CDK12. Involved in the recruitment of transcription elongation and termination factors.	3,5
Ssu72	Ser5P phosphatase, involved in DNA looping.	12
Cstf64	Recognizes the U/GU-rich sequence 30 bp downstream of the poly(A) signal cleavage site.	12
CPSF73	Subunit of the CPSF complex. Involved in mRNA cleavage.	49
CDK9	Kinase, subunit of the P-TEFb heterodimer. Phosphorylates Ser2 of the pol II CTD, NELF-E, and Spt5.	12
CDK12	Kinase interacting with cyclin K. Phosphorylates Ser2 of the pol II CTD.	12
TFIIH?	Protein complex containing CDK7 involved in transcription initiation and termination. Phosphorylates Ser5 and Ser7 of the pol II CTD.	45,46
Mediator?	Protein complex involved in transcriptions regulation and in the formation of enhancer-promoter gene loops.	44,50

Are the checkpoints at the beginning and end of genes connected?

Terminal splice sites and both the AAUAAA of the poly(A) site and the GT/T-rich region downstream, are good candidates for setting up and positioning the poly(A)-associated checkpoint as their recognition is tightly linked to poly(A) site function.³⁴ However, it is still not clear whether premature termination is linked to functional recognition of poly(A) sites.³⁵ To understand the spatial relationship between poly(A) sites and premature termination caused by kinase inhibition, it will be necessary to accurately map both the point of termination and the co-transcriptional usage of poly(A) sites as represented by nuclear RNA rather than total RNA. The newly-developed base-pair resolution techniques for nascent elongation transcript sequencing (NET-seq) and precision nuclear run-on (PRO-Seq) in human cells will help to more accurately map termination.^27,36,37 However, our GRO-seq and ChIP analyses indicate that termination can occur very close to poly(A) sites and perhaps even before the poly(A) site is transcribed, suggesting that pol II might meet a signal or barrier upstream of the poly(A) site.¹² The DNA sequence of the poly(A) site itself could nucleate formation of a complex that extends upstream. Alternatively, recognition of the poly(A) site in a “pioneering” round of transcription could set up the checkpoint in this region. It is well established that pol II regulatory elements widely separated in linear space can interact through looping.³⁸ Several studies have shown that transcription-dependent loops can exist between the promoter and termination region of genes in yeast^39,40 and higher eukaryotes.⁴¹ Thus, looping between the EEC and the poly(A) region could create a barrier for pol II to negotiate or send a signal that a poly(A) site has been reached. However, inhibition of CDK9 can disrupt such loops.⁴¹ In addition, Ssu72, which is a key player in looping between the beginning and end of genes⁴² is lost from the region of the poly(A) site upon CDK9 (or CDK12) inhibition.¹² Therefore, looping between the EEC and the poly(A)-associated checkpoint may be lost when P-TEFb is inhibited. Interaction between the beginning and end of genes could eg supply CDK9 to both the EEC and poly(A)-associated checkpoint at the same time to coordinate their function (Fig. 2). Supporting this notion, CDK9 inhibitors cause loss of CDK9 from the poly(A) site region¹² but not from the EEC.³³

Figure 2.

Are the checkpoints at the beginning and end of genes connected? Left: In normal conditions, the early elongation checkpoint (EEC) and the poly(A)-associated checkpoint could be connected through looping mediated by Ssu72. P-TEFb recruited at the EEC allows the elongating pol II to negotiate both the EEC and the poly(A)-associated checkpoint. This allows transcription past the poly(A) site and the production of a mRNA. Right: In the presence of CDK9 inhibitors, Ssu72 and P-TEFb association with the poly(A) region is lost, suggesting that looping is disrupted when P-TEFb is not active. The inhibition of P-TEFb blocks pol II transcription through the EEC, whereas pol II which has passed the EEC before the addition of the drugs terminates prematurely at the poly(A) site. Red crosses indicate that pol II cannot elongate past this point.

The Med18 subunit of Mediator is present at both the 5′ and 3′ ends in some yeast genes.⁴³ This protein has been shown to recruit CF1 and CPF complexes that are required for 3′-end processing and transcription termination. Recently, the kinase subunit of TFIIH in yeast, Kin28, has also been shown to crosslink to both the 5′ and 3′ ends of genes and has been implicated in transcription termination.⁴⁴ Inhibition of the Kin28 causes a termination defect and the concomitant loss of looping between the ends of genes. In analogy, the human homolog, CDK7, plays an important role in the termination of transcription.⁴⁵ These results raise the possibility that Mediator and CDK7 regulate both gene looping and poly(A)-associated checkpoints in mammalian cells.

The biological utility of a poly(A)-associated checkpoint

The biological significance of poly(A)-associated checkpoints is still unclear but they could be employed to check the quality of the mRNA being produced or mediate responses to developmental and environmental cues (Fig. 3). In normal conditions, after accurate splicing and recognition of a poly(A) site, the elongating pol II can readily negotiate the poly(A)-associated checkpoint and produce an mRNA (Fig. 3A). However, incomplete assembly of the cleavage/polyadenylation complex could trigger premature termination of pol II and degradation of the incorrectly-processed pre-mRNA (Fig. 3B). Likewise, if splicing is inefficient/incorrect the elongating pol II could be held at the poly(A)-associated checkpoint until the mRNA is properly spliced or terminated prematurely (Fig. 3C).

Figure 3.

The biological utility of a poly(A)-associated checkpoint. (A) The early elongation checkpoint regulates the transition to productive elongation whereas the poly(A)-associated checkpoint is quickly negotiated by the pol II if the mRNA quality and the poly(A) site recognition is correct. The negotiation of both checkpoints leads to the production of an mRNA. (B) When a poly(A) site is not correctly recognized, pol II might not be able to negotiate the checkpoint due to the absence of a specific factor(s). The pol II therefore terminates prematurely and the mRNA may be degraded. (C) When the pre-mRNA is not efficiently or properly spliced, the pol II might not be able to negotiate the poly(A)-associated checkpoint until intron splicing is complete or in the case where accurate splicing does not happen, pol II is terminated prematurely and the mRNA may be degraded. (D) In the presence of an environmental or developmental cue, transcription can be stopped at both the EEC and the poly(A)-associated checkpoint simultaneously by inhibiting CDK9 (and CDK12?), abruptly stopping the synthesis of mRNAs. The EEC and poly(A)-associated checkpoints may communicate, eg by looping (Fig. 2). Go sign: pol II can negotiate the early elongation checkpoint or the poly(A)-associated checkpoint; Stop sign: pol II cannot negotiate the checkpoint.

Moreover, the EEC and the poly(A)-associated checkpoint could cooperate to effectively and rapidly regulate both long and short genes in response to environmental signals/developmental cues (Fig. 3D). For example, loss of elongation competence at the EEC and poly(A) site simultaneously provides an efficient way to completely shut-down gene expression, particularly of long genes, where the mRNAs in progress would otherwise be completed.

The EEC,¹ splicing checkpoints,⁴⁶ and the poly(A)-associated checkpoint can therefore all constitute obstacles⁴⁷ that pol II navigates only when conditions are right.

Future prospects

Poly(A)-associated checkpoints provide the potential for powerful, rapidly-responsive control of transcription in response to a range of signals, eg during development where synchronous activation and repression of gene expression is required.² Misregulation of transcription elongation is implicated in developmental defects and cancer.^2,48 Characterization of the structure and function of this novel transcription elongation checkpoint will provide new insights into regulation of gene expression at the level of transcription elongation and how this might be misregulated in human disease.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgments

The authors thank Steve Buratowski and Justyna Zaborowska for helpful discussions and critical reading of the manuscript.