Transcriptional elongation control in developmental gene expression, aging, and disease

SUMMARY

The elongation stage of transcription by RNA polymerase II (Pol II) is central to the regulation of gene expression in response to developmental and environmental cues in metazoan. Dysregulated transcriptional elongation has been associated with developmental defects as well as disease and aging processes. Decades worth of genetic and biochemical studies have painstakingly identified and characterized an ensemble of factors that regulate Pol II elongation. This review summarizes recent findings taking advantage of genetic engineering techniques that probe functions of elongation factors in vivo. We propose a revised model of elongation control in this accelerating field by reconciling contradictory results from the earlier biochemical evidence and the recent in vivo studies. We discuss how elongation factors regulate promoter-proximal Pol II pause release, transcriptional elongation rate and processivity, Pol II stability and RNA processing and how perturbation of these processes is associated with developmental disorders, neurodegenerative disease, cancer and aging.

eTOC blurb

Transcriptional elongation by RNA polymerase II is a critical regulatory step for gene expression and has been linked to the mechanisms underlying human disease and aging processes. This review provides a comprehensive overview of both the current mechanistic understanding and the known biological relevance of Pol II elongation factors.

INTRODUCTION

The first link between transcriptional elongation control and human cancer was identified in 1996, when the Eleven-nineteen Lymphoid Leukemia gene (ELL), a translocation partner of the Mixed Lineage Leukemia gene (MLL), was found to encode a factor controlling elongation by RNA polymerase II (Pol II).¹ This initial report suggested that the elongation stage of transcription is a regulatory step in gene expression and plays a central role in development, and its dysregulation can cause disease pathogenesis, which had not previously been appreciated.¹ Since, it has been demonstrated that the diverse MLL translocation partners found in pediatric leukemias include many subunits of the ELL-containing Super Elongation Complex (SEC), which regulates the release of paused Pol II into productive transcriptional elongation via the activity of the Pol II CTD kinase CDK9 within SEC.^2,3 During the past two decades, these and many other studies have demonstrated the significance of transcriptional elongation control in the regulation of gene expression by defining functional roles for various elongation factors regulating the elongation stage of transcription within physiological contexts relevant to human health and disease. In this review, we provide an up-to-date and comprehensive picture of the mechanisms by which Pol II elongation factors regulate gene expression, as well as their functional roles in development, disease, and as recently been demonstrated in the aging process. These findings further affirm the initial observation that the elongation stage of transcription is a key regulatory step in gene expression.¹

The transcription cycle catalyzed by Pol II is a remarkably conserved mechanism for gene expression within eukaryotes. This cycle begins with the initiation of transcription, followed by promoter-proximal pausing; pause release allows productive elongation and concurrent transcript processing, and the cycle ends with transcription termination. For control of transcription, both the initiation and elongation steps have been shown to be the key regulatory steps. During the elongation step, Pol II traverses genetic distances up to several hundreds of kilobases, and its speed and processivity throughout the genome must be exact for accurate and timely synthesis of a given transcript.

Pol II catalyzes the processive addition of nucleotides to nascent transcripts as it traverses DNA wrapped around histone octamers. Meanwhile, these transcripts also undergo post-transcriptional processing, including 5’ capping and splicing. In addition to these elongation events, Pol II is known to pause at promoter-proximal regions around the first nucleosomes in most metazoans, including human, mouse and fly.^3-5 Paused Pol II is then either released into productive elongation at gene bodies or removed from promoter-proximal regions without production of full-length transcripts.⁴ This promoter-proximal pause/release of Pol II is a fundamental regulatory step for precision both in timing of gene expression and in modulation of expression levels. It is therefore important to understand how elongating Pol II achieves processivity, nucleosomal passage, post-transcriptional processing, and promoter-proximal pause/release to generate properly processed transcript at the right time.

Over the past several decades, a variety of elongation factors have been identified using in vitro transcription assays, genetics, and proteomics approaches.^3-5 As we will discuss in this review, these elongation factors interact with the transcribing Pol II to control its elongation activity, processivity and transcript maturation properties. Notably, the investigation of elongation factors in vivo had been challenging for decades because many of these factors are essential for cellular growth, and their deletion or reduction by shRNA results in the loss of viability. However, due to the recent advent of targeted protein degradation approaches,^6,7 our mechanistic understanding of elongation inside the cells, rather than in the reconstituted system lacking many of the essential factors, has been dramatically advanced. The success of these approaches can be attributed to the rapidity with which transcription factors are depleted, within 30 minutes to a few hours. This rapid depletion allows for the observation of primary phenotypes decoupled from any secondary or indirect effects caused by growth defects upon long-term depletion. Also, when a targeted gene is genetically tagged with a degron peptide, off-target effects can be significantly minimized, allowing for identification of the true, direct target of a given factor. While some of the results obtained using these degradation approaches are consistent with previous findings, there have also been many contradictory observations between reconstituted system in test tubes and what actually happens inside the cells. In this Review, we aim to reconcile these contradictions between in vitro and in vivo studies, providing a new model for the molecular mechanisms of elongation factor function in living cells. In particular, we focus on the elongation factors such as Cyclin-Dependent Kinase 9 (CDK9), Bromodomain and Extraterminal (BET) proteins, Super Elongation Complex (SEC), CDK12/13 kinases, Negative Elongation Factor (NELF), Suppressor of Ty 5 (SPT5), SPT4, SPT6, Pol II-Associated Factor 1 Complex (PAF1C) and Integrator (Figure 1). We also discuss how misregulation of elongation is associated with development, aging and disease pathogenesis.

Figure 1. A graphic summary of Pol II elongation factors.

Factors that regulate transcriptional elongation by Pol II at promoter-proximal regions and gene bodies are shown.

CDK9-CONTAINING POSITIVE ELONGATION FACTORS

Cyclin-dependent kinase 9 (CDK9) and Cyclin T form a complex called Positive Elongation Factor b (P-TEFb) in eukaryotic cells, function of which has been shown to be central to stimulate Pol II elongation genome-wide. Unlike cell cycle-related Cyclins, whose oscillating protein levels regulate the activity of partner kinases, the protein levels of transcription-related Cyclins, including Cyclin “T” (transcriptional), do not oscillate in growing cells. Instead, the CDK9 kinase is regulated via more complex mechanisms that involve the interaction with coactivators, including BRD4 and the SEC family.³ Because CDK9 plays key roles in disease processes, researchers have proposed therapeutic approaches that make use of inhibitors developed to target CDK9 or its coactivators BRD4 and SEC family. In this section, we discuss our current understanding of how CDK9 kinase activity is regulated by BRD4 and SEC and recent advances in the development of CDK9 inhibition strategies for suppressing human diseases including cancer. Functions of the other two transcriptional elongation kinases CDK12 and CDK13 are also discussed. For other aspects of CDK9 regulation including the inactive 7SK snRNP complex, please refer to the recent review.⁸

Identification of CDK9 –

Several early studies of transcription identified the CDK9-Cyclin T complex P-TEFb as a positive regulator of transcriptional elongation. Using in vitro transcription assays, P-TEFb was identified from Drosophila nuclear extract as a target of the pan-CDK inhibitor DRB, an adenosine analog that blocks Pol II elongation.^9-11 The CDK9-Cyclin T complex was also identified as the host cell binding partner of the HIV-1 (human immunodeficiency virus-1) protein Tat, which stimulates Pol II elongation at the viral promoter LTR (long terminal repeat).^11-14 The BUR1 kinase and BUR2 cyclin homologs of CDK9 and Cyclin T in S. Cerevisiae were identified through a yeast genetic screen that isolated transcriptional regulators.¹⁵ While an early study first demonstrated that CDK9 phosphorylates the Pol II CTD,¹⁶ a variety of other CDK9 phosphorylation targets have been identified through phosphoproteomics both in vivo and in vitro.^17-20 CDK9 kinase activity is considered to be essential for Pol II release, as acute inhibition of CDK9 kinase activity blocks the genome-wide release of promoter-proximal Pol II in vivo.^21-23

CDK9 active complex 1: BRD4-CDK9 –

The Bromodomain and extra-terminal domain (BET) protein family consists of BRD2, BRD3, BRD4 and the testis-specific BRDT in mammalian cells. These proteins share two structurally conserved tandem bromodomains at their N-termini that recognize acetylated lysine residues within histone tails.²⁴ Interaction between BRD4 and the CDK9-Cyclin T complex was initially identified through purification of BRD4 from nuclear extracts,²⁵ and this interaction was subsequently found to be required for the stimulation of transcriptional elongation in vitro.²⁶ Biochemical studies demonstrated that the CDK9-Cyclin T complex directly interacts with the conserved C-terminal domain of BRD4 and BRDT, called the P-TEFb-interacting domain (PID),²⁷ and further demonstrated that the PID alone is sufficient to stimulate CDK9 kinase activity in vitro.²⁸ These results suggest a model in which BRD4’s C-terminal PID enhances CDK9 kinase activity, whereas BRD4’s N-terminal bromodomains interact with histone acetylation-enriched chromatin regions, including active promoters and enhancers. BRD4 contains a large intrinsically disordered region between bromodomains and PID.²⁹ The formation of nuclear condensates involving this unstructured region and its function in cancers have been reviewed elsewhere.³⁰

The function of BRD4 bromodomains in transcriptional regulation has been extensively studied for more than a decade now, after two BET bromodomain inhibitor lead compounds (JQ1 and I-BET151) were shown to have anti-cancer activity.^31,32 A knockdown screen targeting chromatin regulators also independently showed that BRD4 depletion results in growth suppression in aggressive leukemia.³³ Despite widespread chromatin localization of BET proteins, BET bromodomain inhibition is effective for downregulation of a small subset of genes, including the oncogene MYC and its target genes.^32-34 This finding, in combination with the BRD4-Mediator interaction,^25,35 led to the identification of super-enhancers, genomic elements that are co-occupied by exceptionally high levels of BRD4 and the Mediator complex.³⁶ BET bromodomain inhibition results in preferential loss of BRD4 at super-enhancers, which contributes to the selective downregulation of the MYC gene.³⁶ Consistent with this result, JQ1 resistance observed in certain cancer cell lines is associated with an enhanced interaction between BRD4 and Mediator that prevents the loss of BRD4 from super-enhancers otherwise induced by JQ1.³⁷

Early studies suggested that the inhibition of BET bromodomain by JQ1 results in increased Pol II pausing at a small subset of genes.^36,38 In contrast, the degradation of BET proteins by dBET6 results in genome-wide induction of pausing.³⁹ This BET degradation phenocopies CDK9 inhibition,^39,40 suggesting that CDK9 regulates Pol II pause/release through BET proteins. Direct and specific rapid degradation of BRD4 confirmed that the pausing phenotype induced by BET degradation is specifically due to loss of BRD4,^41-43 and not BRD2 or BRD3, which lacks interaction with CDK9.⁴² A recent genetic complementation study demonstrated that Pol II pausing induced by BRD4 depletion can be rescued by bromodomain-less, C-terminal fragment of BRD4.⁴⁴ This fragment can interact with the CDK9-CyclinT complex but not acetylated histones.⁴⁴ Thus, BRD4-CDK9 promotes Pol II release independently of chromatin association via binding to acetylated histones. This “acetyl-unbound” layer contrasts the previously described function of “acetyl-bound” BRD4 at super enhancer-controlled genes.³⁶ The acetyl-bound and acetyl-unbound BRD4 are likely to have distinct mechanisms to regulate Pol II elongation. Of note, developing a therapeutic strategy targeting the acetyl-unbound BRD4 will be beneficial when treating cancers that exhibit JQ1 resistance but require BRD4 for their growth. In this strategy other BET proteins, which function at enhancers independently of BRD4,⁴² will be left intact otherwise inhibited by JQ1.

Another BRD4 rapid depletion study also showed that BRD4 is required to recruit the 3’-end processing machinery, including endonuclease CPSF73, to Pol II at 5’ end of genes.⁴³ CPSF73 knockdown has been shown to result in accumulation of promoter-proximal Pol II associated with CTD Ser2 phosphorylation.⁴⁵ These results suggest that BRD4 mediates binding of the 3’-end processing machinery to promoter-proximal Pol II and that the 3’-end processing machinery functions at both the 5’ and 3’ ends of genes.

CDK9 active complex 2: SEC-CDK9 –

Chromosomal translocations involving the MLL gene are found in approximately 10% of leukemic patients; leukemogenic MLL translocations cause fusion of MLL with one of fusion partner genes, including AFF1 (also known as AF4), AF9, ENL and ELL.⁴⁶ Biochemical isolation of these MLL-fusion partner proteins from nuclear extracts led to the identification that many of the MLL translocation partners found on different chromosomes are all part of the Super Elongation Complex (SEC), a large protein complex that contains AFF1, AFF4, ENL/AF9 and ELL1/2 along with CDK9, Cyclin T and other proteins.^47,48 SEC was also independently identified as a large host protein complex interacting with the HIV-1 viral transactivation protein Tat, but Tat is not required for the formation of SEC.^49,50 Because it has strong CDK9 kinase activity with Pol II CTD as a substrate in vitro,^47,51 SEC is considered to be one of the active CDK9-containing complexes apart from BRD4-CDK9. Of note, SEC-CDK9 and BRD4-CDK9 are independent complexes as SEC does not interact with BRD4.⁵⁰ The AF4 family proteins AFF1 and AFF4 interact with other SEC subunits along their structurally disordered axis⁵² and therefore serve as the central scaffolds of SEC. The conserved N-terminal domains of AFF1 and AFF4 directly interact with the CDK9-Cyclin T complex.⁴⁸ The C-terminal domains are involved in AFF1-AFF4 heterodimerization, which is preferred over homodimerization.⁴⁸ The mid-protein regions of AFF1 and AFF4 interact with ENL/AF9 and ELL1/2;^48,52 AFF4 also stabilizes ELL1/2.^47,50 The inhibition of CDK9 kinase activity results in loss of ELL2 recruitment but does not affect AFF4-Cyclin T binding,^49,50 suggesting that CDK9 kinase activity is required for the formation of SEC through stabilization and/or recruitment of ELL2 to AFF4. In vitro, both ELL1 and ELL2 were shown to stimulate the rate of elongation by RNA Pol II and promote synthesis of longer transcripts.^1,53 A structural study showed that the direct interaction of an ELL2 and ELL-associated factor 1 (EAF1) heterodimer with the surface of Pol II contributes to the stabilization of the closed conformation of the Pol II cleft, leading to the allosteric stimulation of elongation.⁵⁴ These results suggest that the AFF1/4 SEC scaffold recruits ELL2-EAF1 and CDK9-Cyclin T modules to Pol II to stimulate Pol II elongation. SEC-PAF1C interaction^49,55 may also contribute to elongation stimulation, but SEC can be recruited to genes in the absence of PAF1.⁵⁶

The other AF4 family proteins AFF2 and AFF3 can also form SEC-like complexes that contain ENL/AF9 and the CDK9-CyclinT kinase module but lack ELL1/2.⁵¹ ELL1 can form another SEC-like complex called Little Elongation Complex (LEC), which consists of the scaffold subunit ICE1 and subunits ICE2 and ZC3H8 but lacks the CDK9-CyclinT kinase module.^57,58 LEC has been shown to stimulate the expression of Pol II-transcribed snRNA genes.^57,58

SEC is a major active CDK9 complex required for rapid transcriptional activation in response to environmental cues. For example, SEC is recruited to stimulate the expression of heat shock-responsive genes upon heat shock stress.^42,47 Depletion of the SEC scaffold subunits AFF1 and/or AFF4 attenuates the rapid induction of heat shock gene expression.^42,47 Similarly, SEC is required for induction of developmental gene expression in response to differentiation stimuli,⁵⁹ for serum-induced immediate early gene activation⁵⁹ and for HIV Tat-induced activation of viral genes.⁵⁰ In contrast, the other major active CDK9 complex, BRD4-CDK9, is dispensable for rapid induction of heat shock genes.⁴² Moreover, because BRD4 and Tat competitively interact with the CDK9-CyclinT kinase module, BRD4 functions to inhibit Tat activation of HIV genes,^26,27,60 which is opposite to SEC’s positive role in Tat activation. These results suggest that SEC uniquely functions to stimulate rapid induction of gene expression such that BRD4-CDK9 does not functionally compensate. While BRD4-CDK9 regulates global release of promoter-proximal Pol II as discussed above, SEC appears to regulate Pol II with gene specificity. For example, BRD4 rapid depletion causes global downregulation of nascent transcripts, which is consistent with genome-wide Pol II pausing.⁴¹ On the other hand, genome-wide studies showed that knockdown of the SEC scaffold subunits AFF1/4 results in downregulation of a specific subset of genes.^51,61 It remains unclear how SEC’s gene specificity is achieved and whether BRD4-CDK9 and SEC have different CDK9 phosphorylation targets. Gene-specific recruitment of SEC via Mediator⁶² and MYC⁶¹ may contribute to the apparent gene specificity of SEC function.

Distinct transcriptional elongation kinases: CDK12 and CDK13 –

In addition to CDK9, the human genome contains two other elongation-related kinase genes: CDK12 and its paralog CDK13. The functions of CDK12 and CDK13 in transcriptional elongation have been extensively studied since their initial identification as CDK-related kinases.^63,64 CDK12 and CDK13 share an exceptionally large size among CDKs and a unique sequence feature. With extended N- and C-termini, the human CDK12 (1,490 aa) and CDK13 (1,512 aa) are approximately 4 times as large as CDK9 (372 aa). The sequences of CDK12 and CDK13 both contain the arginine/serine-rich (RS) domain, which is widely conserved in splicing factors.^63,64 CDK12 and CDK13 localize at nuclear speckles, where splicing factors are predominantly enriched,^63,64 and they physically interact with splicing factors.^65,66 CDK12 and CDK13 also regulate the expression of splicing factor genes.⁶⁵ Depletion of CDK12 or CDK13 results in splicing defects.^65,66 These results suggest that CDK12 and CDK13 function to facilitate splicing not only via direct interaction with splicing factors but also via regulation of expression for splicing factor genes. While CDK9 regulates transcriptional elongation genome-wide, CDK12 shows a unique gene specificity. CDK12 is required for the expression of DNA damage response genes^65,67,68 and CDK12-depleted cells are sensitive to DNA damage agents.⁶⁷ Thus, CDK12 functions to protect cells from genomic instability by facilitating DNA damage response. Inhibition of CDK12/13 kinase activity using the covalent inhibitor THZ531⁶⁹ also results in a gene length-dependent elongation defect, in which long genes are preferentially downregulated due to premature termination that results from premature cleavage and polyadenylation of nascent RNA at the middle of genes.⁷⁰ Interestingly, DNA damage response genes are enriched among these long, CDK12/13-dependent transcripts.⁷⁰ Together, the dependency of long transcripts on CDK12/13 and the enrichment of DNA damage response genes among these transcripts may explain how CDK12/13 achieves its apparent gene specificity. PAF1C-mediated recruitment of CDK12⁷¹ may also be involved in the mechanism of CDK12’s gene specificity.

Immunoprecipitation of Cyclin K from human cell extracts led to the identification of CDK12 and CDK13 as its partner kinases.⁶⁷ While Cyclin K forms the two distinct complexes CDK12/Cyclin K and CDK13/Cyclin K, it does not interact with the other elongation kinase CDK9 in human cells.⁶⁷ This suggests specificity of the CDK/Cyclin interaction, with the following elongation kinase complexes present in vivo: CDK12/Cyclin K, CDK13/Cyclin K and CDK9/Cyclin T. Human CDK12 phosphorylates Ser2 CTD,⁶⁷ and depletion of CDK12 or Cyclin K results in a limited decrease in the bulk levels of Ser2P CTD, but depletion of CDK13 has no effect.^65,67,72 In fly, CDK12 (a homolog of human CDK12/13) also interacts with Cyclin K and phosphorylates Ser2 CTD.⁷² A complementation assay revealed that human CDK9 is a functional homolog of the yeast kinase Bur1, whereas human CDK12/13 is a functional homolog of the yeast kinase Ctk1.^72,73 Ctk1 phosphorylates Ser2 CTD and Ctk1 knockout affects the recruitment of polyadenylation factors in yeast.⁷⁴ In human, CDK12 depletion results in loss of Ser2P towards the 3’ end of genes, with concomitant loss of 3’ end processing factor recruitment.⁷⁵ These results suggest that CDK12 has a conserved function in Ser2 phosphorylation and recruitment of 3’ end processing machinery. Phosphoproteomics analysis using the inhibitor THZ531 revealed that CDK12/13 also phosphorylate splicing and 3’ end processing factors.⁷⁰ This supports the idea that regulation of RNA processing involves CDK12/13-dependent phosphorylation of RNA processing factors.

NELF –– NOT JUST A NEGATIVE ELONGATION FACTOR

NELF has been recognized as a “negative” elongation factor since its biochemical purification following in vitro transcription assay systems by Handa and colleagues.⁷⁶ However, in vivo studies have shown that the depletion of NELF from cells impairs Pol II elongation, suggesting a “positive” elongation factor role for NELF in cells. This discrepancy between in vitro and in vivo studies indicates the need for a careful reconsideration of roles for NELF. In this section, we discuss our current understanding of how NELF regulates Pol II elongation in vitro and in vivo.

The discrepancy between in vitro and in vivo studies regarding NELF’s role in Pol II elongation-

The in vitro inhibitory function of NELF in transcription was first demonstrated through biochemical analyses of reconstituted transcriptional activity on a naked DNA template.⁷⁶ An in vitro structural study suggested that NELF inhibits the catalytic activity of Pol II in an allosteric manner.⁷⁷ The interaction of NELF with Pol II leads to a tilted DNA–RNA hybrid that impairs the addition of nucleotides to nascent transcripts.⁷⁷ A functional connection between the inhibitory function of NELF in vitro and promoter-proximal pausing in vivo has been proposed, as organisms that possess NELF genes show more widespread pausing.⁷⁸

One might expect that NELF depletion in cells would function as though brakes had been disengaged on Pol II, releasing it into gene bodies. However, the tests of this model using knockdown approaches have suggested that NELF may not just act as brakes. Most genes largely retain promoter-proximal Pol II after NELF depletion.^22,79-82 An exception is the hsp70 gene in Drosophila, where promoter-proximal Pol II is lost upon NELF depletion.^83,84 At many genes, NELF-depleted cells exhibit decreased expression.^80,82 Mechanisms underlying these phenotypes in NELF-depleted cells had been unclear. Recently, studies using targeted degradation approaches have provided new insights into the biological role of NELF. Upon rapid depletion of NELF, promoter-proximal Pol II fails to be released, as shown by a reduction of elongating Pol II at gene bodies.^85,86 While promoter-proximal Pol II is retained upon NELF loss, the pause sites are shifted slightly downstream.^85,86 Notably, NELF-depleted cells exhibit inhibition waves of Pol II elongation at gene bodies, representing a block of Pol II release.⁸⁶ Similar inhibition waves have been observed upon the inhibition of CDK9, a positive elongation factor kinase.²¹ We propose that NELF is not just a negative elongation factor, but a “positive” elongation factor that regulates the efficient release of Pol II into gene bodies. Consistent with this idea, the forced addition of NELF to transcribing polymerase does not negatively affect Pol II elongation and does not force pausing, suggesting that pausing of Pol II can take place independently of NELF.⁸⁷

The transition to the downstream pause sites upon NELF loss^85,86 may be a key observation for reconciling the inhibitory function of NELF in vitro with an elongation-allowing mechanism in vivo. In the presence of NELF, Pol II pause sites (denoted the 1^st pause sites) are associated with the entrance to the +1 nucleosome,⁸⁵ which is ~70 bp upstream of the dyad position. This position has also been shown to be a physical barrier to Pol II and was previously interpreted as a cause of Pol II stalling after pause release.⁸⁸ These results suggest that NELF arrests Pol II at the 1^st pause sites, likely through an allosteric inhibition of Pol II.⁷⁷ Consistent with this idea, NELF was shown to be required for accumulation of the 1^st pause site-associated Pol II accumulation upon CDK9 inhibition.^85,89 Secondary pause sites observed in NELF-depleted cells are associated instead with the dyad region of the +1 nucleosome,⁸⁵ which physically blocks Pol II elongation. Disruption of the +1 nucleosome results in the promotion of Pol II release, supporting the idea that the +1 nucleosome functions in Pol II pausing.⁹⁰

Stabilizing 5’ end of transcripts by NELF –

How NELF plays a positive role in elongation is an important open question. Stabilization of nascent transcripts is crucial for proper control of elongation, as degradation of nascent transcripts has been linked to transcription termination.⁹¹ Notably, it has been suggested that XRN2-dependent degradation at promoter-proximal regions triggers premature termination.⁹² NELF is required for the long half-life of promoter-proximal Pol II, suggesting a role for NELF in blocking premature termination.²² Interestingly, a series of studies has suggested a role for NELF in 5’ capping, a co-transcriptional modification of nascent transcripts. Biochemical studies have found that NELF interacts with the Cap-Binding protein Complex (CBC).^93,94 CBC interacts with 5’ capped RNA, stabilizing nascent transcripts by blocking decapping that would lead to degradation from the 5’ end.⁹⁵ A recent study demonstrated that NELF recruits CBC to promoter-proximal regions in vivo.⁸⁵ Upon NELF depletion, the 5’ cap is subsequently lost at a subset of genes.⁸⁵ This coincides with an increased occupancy of the decapping enzyme DCP2 and the 5’-3’ exonuclease XRN2 at promoter-proximal regions.⁸⁵ It appears that NELF stabilizes nascent transcripts via CBC recruitment. The “tentacle” of the NELF-E subunit binds to CBC through the C-terminal region of NELF-E^93,94 but also to promoter-proximal transcripts through the RNA Recognition Motif (RRM) domain of NELF-E.⁹⁶ It is thus possible that the interaction of the NELFF-E RRM with nascent transcripts assists in the recruitment of CBC. Interestingly, this RRM is dispensable for the inhibitory function of NELF in vitro,⁷⁷ suggesting a specific role for the RRM in elongation in vivo, such as CBC recruitment. The NELF-E tentacle is also involved in condensate formation at stress conditions.⁹⁷ This suggest that partitioning of NELF may alter the interaction with CBC and nascent transcripts. Collectively, these studies support the idea that NELF plays a positive role in elongation via CBC recruitment and subsequent stabilization of transcripts, thereby suppressing premature termination and promoting an efficient release of Pol II into gene bodies (Figure 2).

Figure 2. Promoter-proximal regulation by NELF.

(A) NELF recruits CBC to the 5’ end of a nascent transcript. CBC stabilizes a nascent transcript via 1) prevention of decapping by DCP2 and 5’ degradation by XRN2, 2) recruitment of U1 snRNA to block PAS-mediated cleavage by CPSF and CstF, and 3) suppression of Integrator-mediated cleavage. Stabilization of a nascent transcript facilitates efficient release of promoter-proximal Pol II. (B) In the absence of NELF, promoter-proximal Pol II at the +1 nucleosomal dyad is prone to premature termination due to destabilization of a nascent transcript by 1) decapping by DCP2 and 5’ degradation by XRN2, 2) PAS-mediated cleavage by CPSF and CstF, and 3) Integrator-mediated cleavage.

Stabilizing 3’ end of transcripts by NELF–

In addition to 5’ end capping regulation, NELF has also been shown to regulate 3’ end processing. A study in human cells demonstrated that NELF is required for cleavage of the 3’ end of snRNA transcripts by Integrator endonuclease and proper processing without polyadenylation.⁹⁸ Knockdown of NELF resulted in polyadenylation of snRNA transcripts and enhanced recruitment of the cleavage stimulation factor (CstF), which stimulates polyadenylation signal (PAS)-mediated cleavage and polyadenylation.⁹⁸ While NELF was shown to directly interacts with Integrator, NELF was dispensable for the recruitment of Integrator, and vice versa.⁹⁸ These results suggest that NELF facilitates Integrator-mediated 3’ end snRNA processing by preventing CstF recruitment. Notably, CstF recruitment to 5’ end of coding genes was also enhanced upon NELF knockdown,⁹⁸ suggesting that NELF functions to block 3’ end processing at promoter-proximal regions. It remains largely unclear how NELF prevents CstF recruitment, but CBC-NELF interaction might be key to this process. CBC has been shown to recruit the splicing factor U1 snRNP to 5’ splice sites.⁹⁹ Once U1 snRNP recognizes 5’ splice sites, it functions to hinder PAS-mediated cleavage.¹⁰⁰ Thus, CBC recruited by NELF is likely to be associated with inhibition of PAS-mediated cleavage through U1 recognition. This idea is further supported by the findings that NELF loss and U1 inhibition exhibit a similar phenotype of Pol II elongation: NELF loss results in 2nd pause around the +1 nucleosomal dyad,⁸⁵ whereas U1 inhibition also causes accumulation of cleaved transcripts around the +1 nucleosomal dyad.¹⁰¹ Apart from PAS-mediated cleavage, it is also unclear whether Integrator function is regulated by NELF at promoter-proximal regions. Given the remarkably stable state of Pol II pausing,^22,79 NELF might function to inhibit, but not stimulate, Integrator at promoter-proximal regions for stabilization of nascent transcripts. Because the forced addition of NELF to transcribing polymerase does not change the elongation properties of Pol II,⁸⁵ NELF is unlikely to stimulate transcript cleavage that leads to premature termination.

THE UNIVERSAL ELONGATION FACTOR DSIF CONTAINING SPT4/5

The role of the elongation factor SPT5 in eukaryotic transcription has been extensively studied over the last two decades. SPT5 is one of the most widely conserved “universal” elongation factors, as prokaryotic homologs of SPT5 (bacterial NusG and archaeal SPT5) show sequence and functional similarities to eukaryotic SPT5.¹⁰² Early biochemical and genetic studies strongly suggested that SPT5 functions to regulate the elongation rate, promoter-proximal pausing and co-transcriptional processes. Recently, rapid depletion studies have demonstrated that SPT5 also functions to stabilize the RNA Pol II complex. In this section, we discuss various functions of the elongation factor SPT5 in eukaryotic transcription.

Roles for SPT5 in the regulation of promoter-proximal pausing and processive elongation-

More than 25 years ago the elongation factor SPT5 was initially identified and characterized in two independent studies. First, a genetic screening in S. Cerevisiae isolated a series of transcription regulators, called SPT (suppressor of the Ty transposable element) genes, including SPT5 and SPT4.^103,104 Yeast cells with SPT5 mutations or SPT4 gene deletion exhibited sensitivity to the nucleotide-depleting drug 6-azauracil, suggesting a role for SPT5 and SPT4 in transcription elongation in vivo.¹⁰⁴ Second, a biochemical study using human nuclear extracts identified the protein complex DSIF (DRB sensitivity-inducing factor), which consists of the human homologs of yeast SPT5 and SPT4, to be required for Pol II pausing in vitro.¹⁰⁵ However, this study also found that, while DSIF is required for Pol II pausing when CDK9 is inactive, when CDK9 is active DSIF is also required for stimulating the rate and speed of transcription elongation.¹⁰⁵ These genetic and biochemical observations led to the notion that SPT5-containing DSIF is involved in the regulation of at least two distinct transcriptional processes: promoter-proximal pausing and processive elongation.

The formation of the DSIF/NELF/Pol II complex is a characteristic feature of promoter-proximal pausing. Biochemical experiments have shown that NELF preferentially interacts with the preformed DSIF/Pol II complex, rather than with Pol II alone.^106,107 In agreement with this in vitro result, SPT5 was shown to be globally required for recruitment of NELF to the first promoter-proximal pause sites in human cells.¹⁰⁸ A recent cryo-EM structure has revealed that formation of the DSIF/NELF/Pol II complex involves direct interaction between the unstructured C-terminal “tentacle” of NELFA and DSIF on the surface of Pol II.⁷⁷ The significance of this interaction is supported by the finding that deletion of the NELFA tentacle results in loss of Pol II pausing in vitro,⁷⁷ although whether or not the NELFA tentacle functions in this manner in vivo remains to be answered. Collectively, these observations suggest that SPT5-containing DSIF facilitates the formation of the DSIF/NELF/Pol II complex at promoter-proximal regions.

Efficient gene transcription requires proper regulation of the rate of transcript elongation. SPT5 boosts the rate of elongation on naked template DNA in vitro.^105,109 The dual interactions of SPT5 observed with both the DNA upstream of transcribing Pol II and the exiting RNA transcript^110,111 suggest that SPT5 maintains the proper relative orientations of DNA and RNA, thereby facilitating productive elongation. SPT5 also colocalizes with actively transcribed genes and with high levels of the associated Pol II CTD phospho-Ser2 modification in cells,^112-114 supporting the idea that SPT5 similarly regulates the elongation rate in vivo. Consistent with this, a study in S. pombe has shown that rapid depletion of SPT5 results in a globally reduced elongation rate and decreased levels of mRNA expression.¹¹⁵ Intriguingly, SPT5 depletion also increased the levels of antisense transcripts due to the activation of non-canonical, antisense TSS proximal to canonical TSS,¹¹⁵ suggesting that SPT5 not only stimulates the rate of elongation but also suppresses antisense transcription in S. pombe.

A newly recognized, conserved role for SPT5 in Pol II stabilization-

Rapid depletion strategies have delivered a breakthrough in our understanding of how SPT5 functions in mammalian cells. Recent work revealed that SPT5 loss results in degradation of the largest Pol II subunit RPB1,^108,116 for which the catalytic activity of the unfoldase VCP/p97 is required.¹⁰⁸ Because SPT5 functions to stabilize Pol II, it was crucial to determine whether transcriptional defects observed in SPT5-depleted cells result from Pol II degradation directly or from degradation-independent changes in Pol II function. Blocking Pol II degradation with the VCP inhibitor NMS-873 revealed a significant reduction in the elongation rate upon SPT5 depletion, decoupled from the effects of Pol II destabilization,¹⁰⁸ suggesting that in addition to stabilizing Pol II, SPT5 also independently functions to stimulate its elongation rate in vivo. Because SPT5 depletion from the nucleus also causes VCP-dependent degradation of Pol II in S. Cerevisiae,¹⁰⁸ SPT5’s function in stabilizing Pol II may be broadly conserved in eukaryotic cells.

In human cells, the E3 ubiquitin ligase Cullin 3 (CUL3) is required for Pol II degradation upon SPT5 depletion.¹⁰⁸ CUL3, which is localized in both cytoplasm and nucleus, is recruited to chromatin and interacts with Pol II in the absence of SPT5,¹⁰⁸ suggesting that SPT5 prevents CUL3 interaction with Pol II. Direct substrates of the CUL3 ubiquitin ligase and the mechanism by which CUL3 recognizes the SPT5-depleted Pol II complex remain to be determined. Transcription-coupled DNA repair and subsequent transcriptional recovery involve ubiquitylation and degradation of the largest Pol II subunit RPB1, mediated by Cullin-RING ubiquitin ligases in mammalian cells.^117,118 While both Cullin ligases including Cullin 5 (CUL5) and non-Cullin ligases including NEDD4 have been reported to be responsible for RPB1 ubiquitylation in mammalian cells,¹¹⁹ a biochemical study showed that the Cullin 4 (CUL4) ubiquitin ligase ubiquitylates the RPB1 subunit of the Pol II elongation complex via its substrate receptor CSA (Cockayne syndrome A).¹²⁰ Notably, neither CUL5 nor NEDD4 was required for RPB1 degradation in SPT5-depleted cells,¹⁰⁸ suggesting a specific role for CUL3 in Pol II degradation in the absence of SPT5. How each ubiquitin ligase contributes to Pol II degradation remains an open question. Intriguingly, structural studies showed that SPT5 is displaced by CSB (Cockayne syndrome B) upon DNA damage,^120,121 suggesting a physiological significance for SPT5 removal from Pol II during transcription-coupled DNA repair. However, whether SPT5 depletion without DNA damage results in CSB recruitment remains unclear. CDK9 kinase activity is also required for Pol II degradation upon SPT5 depletion,¹⁰⁸ but no known CDK9 activators appear to regulate this activity: neither degradation of BRD4 by the PROTAC dBET6 nor inhibition of the SEC by KL2 impedes Pol II degradation upon SPT5 depletion.¹⁰⁸ This result suggests either that CDK9 mediates Pol II degradation without any activators or that it uses uncharacterized activators to trigger Pol II degradation. Consistent with this result, early studies showed that in vitro ubiquitylation of Pol II requires nuclear kinase activity and Pol II CTD.^122,123

Phosphorylation regulates SPT5’s functions-

Given that SPT5 functions in multiple steps of Pol II elongation, it is important to understand how these distinct functions are regulated. It has been shown that phosphorylation of the threonine residues within SPT5’s C-terminal repeat regions (CTR) by CDK9 is required for SPT5 to stimulate the elongation rate both in vitro and in vivo.¹²⁴ To understand the function of SPT5 phosphorylation in living cells, it is useful to deplete wildtype SPT5 and complement with a transgene containing either wildtype SPT5 or SPT5 mutants in which the CTR threonine residues are substituted with alanine (CTR-TA) or glutamate (CTR-TE, a phospho-mimic). Human cells expressing the SPT5 CTR-TE mutant exhibit an increased release of promoter-proximal Pol II and increased levels of mRNA expression at thousands of genes,¹²⁵ supporting the model that phosphorylation of its CTR threonine residues switches SPT5 from a pausing factor to a productive elongation factor.¹²⁴ In contrast, cells expressing the SPT5 CTR-TA mutant exhibit normal Pol II distribution around the 5’ end of genes that is nearly identical to the distribution seen upon WT SPT5 expression,^116,125 suggesting that SPT5 CTR phosphorylation is not required for Pol II release in normal conditions but rather promotes Pol II release upon signaling induced by extracellular stimuli including EGF treatment.¹²⁴ Notably, CTR-TA mutations also result in early termination at the 3’ end of genes.^116,125 The levels of endogenous SPT5 CTR phosphorylation diminish downstream of poly A sites, and dephosphorylation of the CTR by PP1 phosphatase is required for proper termination in S. pombe and human cells.^113,126 These observations suggest that 1) in response to extracellular stimulus, SPT5 CTR phosphorylation at 5’ end of genes promotes the release of Pol II into productive elongation, 2) increased levels of CTR phosphorylation within gene bodies may stimulate the elongation rate and 3) CTR dephosphorylation at the 3’ end of genes slows down Pol II, facilitating termination. It has also been shown that interaction between the SPT5 CTR and the RNA capping enzyme increases capping efficiency and that CTR phosphorylation inhibits this interaction in vitro.^127-131 SPT5 CTR is positioned close to the RNA transcript exiting Pol II,^110,111 supporting the idea that the capping enzyme contacts the 5’ end of newly synthesized RNA via interaction with the SPT5 CTR and that CTR phosphorylation releases the capping enzyme from the RNA exit tunnel to facilitate efficient release of Pol II into productive elongation (Figure 3).

Figure 3. Functions of distinct SPT5 phosphorylation sites during transcriptional elongation.

Location of the KOWw4-5 linker and CTR phosphorylation sites are indicated in the 3D structure model of the Pol II/SPT4/SPT5/NELF complex (PDB: 6GML⁷⁷). Both the KOWw4-5 linker and CTR domains are positioned close to the nascent RNA exiting Pol II. The Pol II elongation-related phenotypes of phospho-mimic/phospho-deficient SPT5 mutants in these phosphorylation sites are also indicated.

CDK9 phosphorylation of the serine residues within the SPT5 KOWx4-KOW5 linker domain²⁰ has also been shown to regulate Pol II elongation at the 5’ end of genes.^116,125 Human cells with the SPT5 S666A mutation, where serine 666 within the KOWx4-KOW5 linker domain is substituted with alanine, exhibit high levels of promoter-proximal Pol II.^116,125 Notably, the increased Pol II release in CTR-TE mutants can be rescued by this phospho-deficient KOWx4-KOW5 linker mutation, resulting in the accumulation of paused Pol II.¹²⁵ These observations indicate that phosphorylation of the KOWx4-KOW5 linker is essential for Pol II release and that CTR phosphorylation may function to stimulate phosphorylation of the KOWx4-KOW5 linker in turn to increase the rate of Pol II release. Integrator-PP2A phosphatase has been shown to dephosphorylate phospho-S666 both in vitro and in vivo,¹³² suggesting that a balance between CDK9 phosphorylation and Integrator-PP2A dephosphorylation of the SPT5 KOWx4-KOW5 linker domain determines whether Pol II is ultimately paused or released at promoter-proximal regions. However, whether SPT5 phosphorylation allosterically affects Pol II processivity or whether it may recruit other elongation factors to modulate Pol II pause-release remains unclear.

THE POLYMERASE ASSOCIATED FACTOR, PAF1 COMPLEX

The elongation factor PAF1 complex (PAF1C), which is associated with both developmental processes and disease states, regulates multiple aspects of Pol II elongation including promoter-proximal pausing, the rate of elongation and 3’ end co-transcriptional processes. PAF1C was initially identified by affinity purification of Pol II-interacting proteins from cell extracts and its composition was characterized by biochemical experiments, from which we know that PAF1C consists of PAF1, CDC73, LEO1, CTR9, RTF1^133-136 and a higher eukaryotic-specific subunit WDR61/SKI8.¹³⁷ Cancer genetic studies also independently identified the PAF1 and CDC73 genes. The oncogene encoding PAF1/PD2 is located in an amplified chromosomal region (19q13 amplicon) that is frequently observed in pancreatic cancers.¹³⁸ The tumor suppressor gene encoding CDC73/Parafibromin/HRPT2 is associated with hyperparathyroidism–jaw tumor syndrome.¹³⁹ In this section, we discuss the functions of metazoan PAF1C in the regulation of Pol II elongation. Please refer to the recent review¹⁴⁰ for the functions of yeast PAF1C and for PAF1’s functions in other regulatory processes including 3’ end processing.

Histone modifications mediated by PAF1C-

PAF1C has been shown to stimulate the transcription-coupled post-translational modification of histones. Early yeast studies revealed that PAF1C is required for monoubiquitination at histone H2B lysine 123 (K123ub), an epigenetic mark associated with active transcription.^141-143 In yeast, PAF1C stimulates the E2 ubiquitin-conjugating enzyme Rad6 and the E3 ubiquitin ligase Bre1, which are both responsible for deposition of H2B K123ub.^141,144-148 PAF1C/Rad6/Bre1-mediated deposition of H2B K123ub was further shown to be required for H3 lysine 4 (K4) methylation by COMPASS and for H3 lysine 79 (K79) methylation by Dot1, in a process of modification-dependent modification called histone cross talk.^149-153 H2B monoubiquitination recruits COMPASS to chromatin to promote H3 K4 methylation¹⁵⁴ and stimulates the methyltransferase activity of Dot1L.¹⁵⁵ Yeast studies also showed that the BUR1/BUR2 kinase complex (yeast homolog of CDK9/Cyclin T) is required for H2B K123ub and H3 methylation, via recruitment of PAF1C.^156,157 Human PAF1C was later shown to mediate deposition of H2B K120ub (equivalent to yeast K123ub), catalyzed by the E2 enzyme RAD6 and the E3 ligase BRE1A/B,^158-162 suggesting a conserved role for PAF1C in stimulating transcription-coupled histone modifications.

PAF1C regulates promoter-proximal pausing-

Initial evidence in human cells suggested that PAF1C can potentially inhibit and/or stimulate Pol II release at promoter-proximal regions.^56,71 The inhibitory role for PAF1 in Pol II release has been strongly supported by various in vivo findings. A key finding is that rapid depletion of PAF1 in human cells results in an increase in Pol II occupancy at gene bodies and, importantly, an increase in the levels of mRNA transcripts from short genes.^87,163,164 While the accumulation of Pol II downstream of TSS could be explained by the reduced elongation rate in PAF1-depleted cells,¹⁶⁵ the elevated polyadenylated mRNA levels in short genes^87,164 strongly suggest the induction of Pol II release upon PAF1 depletion. These observations support the notion that PAF1 depletion causes both release of paused Pol II and defects in processive transcription elongation. Indeed, PAF1 depletion results in premature termination due to loss of processivity, especially at > 30 kb-long genes,¹⁶⁵ consistent with the observation that PAF1 loss leads to reduction of mRNA transcripts from long genes.^87,164 The induction of Pol II release in PAF1-depleted cells was shown to involve stimulation of the CDK9-containing complex SEC⁵⁶ and activation of enhancer transcription,¹⁶³ but PAF1C also appears to inhibit Pol II release through recruitment of Integrator-PP2A phosphatase.¹⁶⁴ The loss of Integrator-PP2A recruitment in PAF1-depleted cells is correlated with the hyperphosphorylation of SPT5¹⁶⁴ (Figure 4). Phosphorylation of SPT5 S666 by CDK9, which is required for Pol II release in human cells,^116,125 can be removed by Integrator-PP2A both in vivo and in vitro.¹³² SPT5 S666 dephosphorylation by Integrator-PP2A may play a role in the PAF1C-mediated inhibition of Pol II release (Figure 4; please also refer to the SPT5 section of this review). An alternative possibility is that the endonuclease activity of Integrator is involved in this process. Pol II release in PAF1-depleted cells could then be explained by lack of endonuclease-mediated transcription attenuation due to loss of Integrator recruitment. In support of a role for the PAF1C in negatively regulating Pol II pause/release, it was demonstrated that PAF1C transiently forms a PAF1C-MYC complex and this complex inhibit Pol II release at MYC-targeted genes, and depletion of the PAF1C subunit CDC73 stimulates the expression of genes targeted by the transcription factor MYC.¹⁶⁶ Furthermore, genetic evidence in zebrafish also supports the idea that PAF1C negatively regulates release of Pol II. A loss of erythroid gene expression observed in the zebrafish tif1γ mutant moonshine was rescued by depletion of PAF1C subunits.¹⁶⁷ Notably, the foggy/spt5 V1012D mutation, which exhibits induced release of Pol II with normal elongation rates in vitro, also rescued moonshine.¹⁶⁷ These results in zebrafish suggest that depletion of PAF1C results in induced release of Pol II, thereby restoring lost erythroid gene expression in moonshine. Finally, depletion of SPT6 causes loss of PAF1C recruitment and stimulates Pol II release in human cells.⁸⁷

Figure 4. PAF1 regulation in promoter-proximal pause-release.

(A) With PAF1C, phospho-SPT5 is slowly increased because PAF1C recruits Integrator-PP2A to counteract SPT5 phosphorylation by CDK9, which is required for release of promoter-proximal Pol II. (B) Without PAF1C, phospho-SPT5 is rapidly increased due to loss of Integrator recruitment, resulting in stimulation of Pol II release. Integrator-PP2A can associate with Pol II through CTD binding even when PAF1C is lost (indicated as the dashed line). After Pol II release, PAF1C provides high processivity and stimulates elongation rate. Without PAF1, Pol II shows low processivity and slow elongation.

Accumulation of promoter-proximal Pol II observed in PAF1-depleted cells can also be explained by Pol II pausing, suggesting a stimulatory role for PAF1 in Pol II release.⁷¹ This stimulation model was strongly supported by the finding that PAF1C is required for CDK9-mediated Pol II release in vitro.²⁰ It remains unclear how PAF1C stimulates Pol II release, but PAF1C recruitment or conformational changes on Pol II may be an important mechanism for this stimulation. The recruitment of PAF1C to Pol II requires the removal of NELF, as PAF1C and NELF are sterically incompatible with simultaneous binding to Pol II.²⁰ The removal of NELF from Pol II results in Pol II release in vitro.⁷⁶ Therefore, the displacement of NELF by PAF1C may be a part of mechanism of Pol II release in vitro. However, it should be noted that NELF depletion in human cells is not sufficient for Pol II release⁸⁵ and that NELF-bound Pol II can also be released into gene bodies.⁸⁷

The key question here is what causes accumulation of Pol II downstream of TSS upon PAF1 depletion. Pol II release, elongation defects, or Pol II pausing may each be a cause of this accumulation, as we discussed above. Importantly, these mechanisms are not mutually exclusive; for example, Pol II release and elongation defects can occur simultaneously. Additionally, the balance of kinase and phosphatase activities appears to be a crucial aspect of the PAF1C-dependent mechanism (Figure 4). Nucleosomes, chromatin remodeling and co-transcriptional processes may also be involved. Therefore, dissecting the functions of PAF1C inside the cells within the complex context of transcriptional elongation through native chromatin will be an important direction toward a better understanding of pause/release mechanisms.

PAF1C regulates the rate of elongation-

A large number of in vitro and in vivo studies have shown that PAF1C positively regulates elongation by stimulating the rate of elongation and maintaining processivity. In vitro, PAF1C is required for promoting elongation on a biochemically prepared chromatin template.¹⁶⁸ This function is distinct from the earlier described histone ubiquitylation- and methylation-dependent chromatin transcription activities of PAF1C,¹⁶⁸ suggesting that PAF1C directly regulates Pol II processivity without modulating histone modifications. The dissociable PAF1C subunit RTF1^137,169 allosterically stimulates the elongation rate on a naked DNA template in vitro.¹⁷⁰ The localization pattern of RTF1within gene bodies is distinct from that of other PAF1C subunits: RTF1 is dissociated from chromatin earlier than other PAF1C subunits towards the 3’ end of genes,⁸⁷ suggesting an RTF1-specific role in the regulation of the elongation rate within gene bodies. Depletion of either PAF1 or RTF1 in mammalian cells significantly lowers the rate of elongation.^165,171 The slowed Pol II found in PAF1-depleted cells also exhibits premature termination at the middle of genes,^164,165 which indicates low processivity. These observations all suggest that PAF1C is required to complete elongation in long genes. Consistent with this notion, rapid depletion of PAF1 in human cells results in the specific downregulation of mRNA transcribed from long genes.^87,164 It should be noted that, while promoter-proximal Pol II is globally released upon PAF1 depletion in these studies, whether this release increases the levels of mRNA transcribed from a gene depends on its length, as mRNA levels are increased for short genes but decreased for long genes.^87,164 Given that elongation defects can depend on gene length in this way, researchers need to be cautious when analyzing genome-wide Pol II distribution data using “meta-gene”, which assumes a single hypothetical gene length. Instead, an analysis using actual gene lengths can provide more informative and reproducible conclusions. Such analyses also help to distinguish the functional regulation of elongation rate from that of Pol II pause-release.

Enhancer RNAs and promoter upstream transcripts-

In addition to its role in the transcription of coding genes, pause/release during elongation is involved in the transcription of enhancer RNAs (eRNAs) and promoter upstream transcripts (PROMPTs) by Pol II.¹⁷² These transcripts are short and unstable compared with coding transcripts. PAF1C was shown to inhibit activation of enhancer transcription in human cells.¹⁶³ A recent study revealed that PAF1C also functions in the transcriptional termination eRNAs and PROMPTs.¹⁷³ PAF1 depletion in mouse cells results in the accumulation of length-extended eRNAs and PROMPTs and the loss of Integrator recruitment to these transcription units.¹⁷³ These results suggest that PAF1C promotes 3’ end processing of eRNAs and PROMPTs by recruiting Integrator endonuclease activity.

PAF1C interaction with other factors-

PAF1C has been shown to physically interact with a large number of proteins. As we discussed above, interaction of PAF1C and Integrator facilitates not only the regulation of promoter-proximal pausing through Integrator-PP2A phosphatase activity,¹⁶⁴ but also the cleavage of non-coding RNA transcripts through the endonuclease activity of Integrator.¹⁷³ In addition, transcription through nucleosomes involves interaction between the PAF1C subunit RTF1 and the FACT complex in vitro,¹⁷⁴ although the functional significance of this interaction remains unclear.

Of note, little data is available regarding the function of the higher eukaryote-specific PAF1C subunit WDR61/SKI8 in transcriptional regulation. Along with SKI2 and SKI3, WDR61/SKI8 forms the superkiller (SKI) complex, a conserved cytoplasmic cofactor of the exosomal RNA-degradation machinery and SKI2 helicase subunit has been shown to channel RNA into the exosome, thereby stimulating exosomal exoribonuclease activity.^175,176 Although PAF1C-dependent recruitment of SKI3 and WDR61/SKI8 to chromatin has been reported,¹³⁷ whether the WDR61/SKI8 subunit of PAF1C functions to facilitate exosomal activity via the SKI complex will require further investigation. There is a notable difference in the stoichiometry of WDR61/SKI8 relative to other subunits in PAF1C and the SKI complex. Pol II-bound PAF1C contains one copy of each individual subunit, including WDR61/SKI8, whereas SKI is a tetrameric complex with 1:1:2 stoichiometry between SKI2:SKI3:WDR61/SKI8.¹⁷⁷ Structural studies of Pol II-bound PAF1C revealed that CTR9 within PAF1C interacts with the top surface of the WDR61/SKI8 beta-propeller structure,^20,170 the same surface that interacts with SKI3 within the SKI complex.^175,176 Though WDR61/SKI8 participates in both the PAF1C and SKI complexes, these structural observations support a model in which the participation of WDR61/SKI8 in either complex is mutually exclusive.

PAF1C recruitment-

Recruitment of PAF1C to Pol II has been shown to require the presence of another elongation factor, SPT6, in living cells.^87,178 Rapid depletion of SPT6 results in the loss of PAF1C recruitment in yeast and human cells,^87,178 suggesting a conserved mechanism for PAF1C recruitment by SPT6. The interaction between the PAF1C subunit CDC73 and SPT6 is required to direct PAF1C to Pol II in S. Cerevisiae,¹⁷⁸ and the structurally determined interaction between CTR9 and SPT6²⁰ may also be involved in PAF1C recruitment. As SPT6 itself is recruited to Pol II in a CDK9 dependent manner (please refer to the SPT6 section of this review), PAF1C recruitment must therefore also be regulated by CDK9. Consistently, CDK9 has been shown to be required for PAF1C recruitment in human cells⁷¹ and for the replacement of PAF1C with NELF in vitro.²⁰ SPT6 is required for PAF1C’s function in pause/release, elongation rate modulation and regulation of chromatin modifications.^87,178 Thus, recruitment of PAF1C by SPT6 and CDK9 is a key initializing step required for PAF1C to functionally regulate multiple steps of the Pol II elongation process.

MULTIPLE FUNCTIONS OF ELONGATIONG FACTOR SPT6

Among the known transcription elongation factors, SPT6 is one of the largest polypeptides (~200 kDa in human), comprising multiple domains, with many protein interactants identified. Here we summarize the multiple functions of SPT6 in transcription elongation.

Identification of SPT6 as an elongation factor –

The SPT6 gene was initially identified through yeast genetic screening by Fred Winston and colleagues.^103,179 This gene is widely conserved among eukaryotic cells, from yeast to human. In yeast, spt6 mutant cells are sensitive to the nucleotide-depleting agent 6-azauracil and show conditional lethality upon gene deletion of TFIIS, suggesting a role for SPT6 in transcription elongation.¹⁰⁴ Consistently, SPT6 localizes to active transcription sites of Pol II in fly.¹¹² Recent studies utilizing rapid depletion of SPT6 have confirmed that SPT6 regulates transcription elongation genome-wide in human cells.^87,171,180 SPT6 has been shown to regulate the rate of elongation both in vitro¹⁸¹ and in vivo.^180,182

Function of the SPT6 C-terminal region –

The C-terminal region of SPT6 contains the tandem SH2 (tSH2) domain, which is required for the SPT6-Pol II interaction.^183,184 The SPT6 tSH2 domain directly binds to the Pol II CTD linker region that tethers the CTD to the Pol II core.^20,185 Phosphorylation of the CTD linker significantly increases the affinity for the SPT6 tSH2 domain in vitro.^20,185 These results suggest that the phosphorylation of the CTD linker by P-TEFb^20,186 or other kinases is likely to stimulate the SPT6-Pol II interaction. Interestingly, deletion or mutation of the SPT6 tSH2 domain causes only a partial reduction of SPT6 occupancy at Pol II-transcribed genes in yeast.^185,187-189 Thus, there is a tSH2-independent mechanism of SPT6 recruitment to gene regions.

Function of the SPT6 N-terminal region –

The N-terminal region of SPT6 has been shown to interact with IWS1/Spn1.^{130,183,190-193} IWS1 is conserved among eukaryotic cells and essential for viability.¹⁹⁴ In addition to binding the RNA export factor REF/Aly Aly,¹⁸³ it forms a multivalent interaction with transcription elongation regulators including ELOA, PPP1R10, LEDGF, HRP2 and TFIIS through conserved modules of disordered protein interaction,^195,196 and a mutation of IWS1 that disrupts binding to LEDGF and HRP2 causes defects in transcription elongation.¹⁹⁵ The IWS1-Pol II interaction is also dependent on SPT6,¹⁹⁷ suggesting that SPT6 recruits IWS1 to Pol II. These results suggest that the SPT6-IWS1 interaction provides a central hub for recruiting a host of factors regulating transcription elongation and mRNA transport (Figure 5). A recent study revealed that SPT6, IWS1, PAF1C and Pol II form a nuclear condensate that facilitates gene activation,¹⁹⁸ suggesting that functions of SPT6-IWS1 hub involve spatial partitioning of elongation factors.

Figure 5. SPT6 recruits numerous elongation factors to Pol II.

(Left) The interaction between the SPT6 C-terminal tSH2 domain and the phosphorylated Pol II CTD linker domain is indicated in the 3D structure model of the Pol II/DSIF/PAF1C/SPT6 complex (PDB: 6GMH²⁰). (Right) IWS1 and the nucleosome are competitively associated with the N-terminal domain of SPT6. IWS1 interacts with many factors including LEDGF, TFIIS, ELOA, and HRP2. Dashed lines indicate competitive interactions of IWS1 with these factors. These interactions may serve the numerous functions of SPT6 to prevent cryptic initiation and promote productive elongation.

In addition to IWS1, the N-terminal region of SPT6 also binds to histones.^190,199,200 Interactions between SPT6 and either histones or IWS1 are mutually exclusive,¹⁹⁰ implicating a regulatory mechanism of factor swapping between IWS1 and histones at the SPT6 N-terminal region. The SPT6-histone interaction has been associated with the function of SPT6 in repressing cryptic initiation at intragenic regions in yeast.^201-206 Further analysis in human cells has shown that long-term (but not rapid) depletion of SPT6 results in cryptic initiation.¹⁸⁰ When cryptic initiation occurs, intragenic binding of TFIIB and TBP increases and nucleosomal positioning patterns are altered.^180,202 Interestingly, mutations of the SPT6 tSH2 domain that disrupt binding to Pol II also cause cryptic initiation in yeast.¹⁸⁵ Thus, repression of cryptic intragenic initiation requires both SPT6-histone and SPT6-Pol II interactions. SPT6 also appears to have histone-related roles beyond transcription, maintaining the centromeric histone variant CENP-A through their direct interaction in flies.²⁰⁷

Promoter-proximal pause/release and PAF1C recruitment by SPT6 –

Recently, SPT6 was found to regulate promoter-proximal pause/release through the recruitment of the PAF1 complex (PAF1C) to Pol II.⁸⁷ Rapid depletion of SPT6 in human cells leads to the release of Pol II from promoter-proximal regions and a defect in PAF1C recruitment to Pol II.⁸⁷ This result is consistent with the previous finding that PAF1 loss leads to Pol II release.^56,163 Interestingly, loss of either SPT6 or PAF1 also causes an observable elongation defect in long genes.^87,164,165 When PAF1C recruitment is impaired, NELF remains bound to Pol II at gene bodies.⁸⁷ Thus, the swap of NELF for PAF1C by SPT6 is important for productive elongation in long genes. This result also suggests that NELF-bound Pol II is transcription-competent, which supports the idea that NELF is a positive elongation factor.

Processivity and 3’ end control by SPT6, PAF1C and CDK12/13 –

Elongation processivity, which is an ability of Pol II to complete transcription for the entire gene length, and 3’ end RNA processing are functionally coupled. This is because processivity is substantially decreased when the nascent RNA is prematurely cleaved and polyadenylated via 3’ end processing factors, resulting in premature termination.²⁰⁸ As we discussed earlier, generation of long transcripts requires PAF1C, SPT6 and CDK12/13 kinases, which strongly suggests the role for these factors in Pol II processivity and 3’ end processing. It remains unclear how PAF1C, SPT6 and CDK12/13 cooperatively regulate these processes. SPT6 recruits PAF1C to Pol II,^87,178 while PAF1C is required for CDK12 recruitment.⁷¹ Interestingly, CDK12 inhibition using an analog-sensitive (as) mutant results in loss of SPT6 and PAF1C.²⁰⁹ A reduction of Ser2P levels upon CDK12-as inhibition²⁰⁹ may contribute to loss of SPT6, as SPT6 preferentially interacts with Ser2P CTD.¹⁸³ These observations raise the possibility that CDK12 plays a key role in a positive feedback system that facilitates processivity, where SPT6 and PAF1C recruits CDK12 and CDK12 supports recruitment of SPT6 and PAF1C. Likewise, regulation of the rate of elongation (or velocity), which requires at least SPT6 and PAF1C, may also involve such cooperative mechanism. This idea is consistent with the finding that SPT6 depletion results in decreased rate of elongation and concomitant loss of processivity.¹⁸⁰

INTEGRATOR: A MULTIFUNCTIONAL REGULATOR OF POL II TRANSCRIOTION

Integrator is a gigantic multisubunit complex that directly interacts with transcribing Pol II. Since its biochemical identification by Ramin Shiekhattar and colleagues,²¹⁰ Integrator has been shown to play various key roles in regulating transcription and RNA processing. Like NELF, Integrator is evolutionarily conserved only in metazoan.²¹⁰ In this chapter, we discuss mechanisms of Pol II elongation regulated by two catalytic modules of Integrator: the RNA endonuclease module and the protein phosphatase module. For details on the structural insights and other functional modules, please refer to the recent reviews.^211,212

Integrator-Pol II CTD interaction –

Integrator has a high affinity for the Pol II CTD. While Integrator interacts with the unmodified CTD,^210,213 CTD phosphorylation at Tyr1, Ser2, and/or Ser7 strengthens this interaction.^214-216 In addition, Integrator also interacts with NELF, SPT5, and the surface of Pol II,^98,217,218 forming a large, promoter-proximal paused Pol II/Integrator complex. Consistent with these lines of structural and biochemical evidence, it has been shown that Integrator, Pol II, NELF and SPT5 all co-localize at promoter-proximal regions.^218-220 Integrator is also recruited to gene bodies²²⁰ via the physical interaction with PAF1C.^164,173 SPT6, which recruits PAF1C to Pol II,^87,178 was also shown to be required for Integrator recruitment to genes.²²¹ These results suggest that (1) at promoter-proximal regions, Integrator is associated with NELF, SPT5, Pol II surface and CTD, (2) in gene bodies, after NELF is replaced with PAF1, Integrator remains associated with Pol II through the CTD and PAF1C. The Integrator’s high affinity to the CTD^210,213-215 may allow for retention of Integrator at Pol II when the NELF-PAF1 exchange takes place.

RNA Endonuclease module of Integrator –

The Integrator endonuclease module consists of INTS4, INTS9 and the catalytic subunit INTS11.^210,222 The catalytically-deficient mutation of INTS11 (E203Q in human) causes various defects in 3’ end processing of transcripts, including U small nuclear RNA,²¹⁰ mRNA,^223,224 enhancer RNA²²⁵ and microRNA.²²⁶ Thus, the endonuclease activity is broadly required for 3’ end processing. Notably, the INTS11 E203Q mutation results in readthrough transcription at 3’ end of mRNA genes.^223,224 This readthrough yields Downstream-of-Gene (DoG)-containing transcript, a type of long non-coding RNA associated with human diseases. For the details of DoG function, please refer to the recent review.²²⁷

Integrator also regulates transcription elongation at 5’ end of protein-coding genes via its endonuclease activity. When the catalytic subunit INTS11 is depleted or mutated, the level of promoter-proximal Pol II is increased.^219,228-231 Two models are widely appreciated to explain how Integrator controls Pol II pause/release. The turnover model proposes that Integrator can sense elongation-deficient Pol II and remove it from promoters, thereby recruiting an elongation-competent Pol II replacement and facilitating the transition of Pol II to productive elongation.^219,220 Direct evidence for this model came from the fact that Integrator is required not only for maintaining steady-state transcription^219,232 but also for rapidly upregulating gene expression in response to the acute stimuli of growth factor EGF or heat shock.²²⁰ In addition, Integrator interacts with the super elongation complex (SEC),²²⁰ which is also required for rapid gene activation.^42,47 Consistent with these results, rapid depletion of INTS11 was recently shown to induce pausing and reduce mRNA synthesis.²³⁰ Thus, in the turnover model Integrator positively regulates gene expression by providing quality control for the Pol II elongation machinery. However, how Integrator recognizes an elongation-deficient Pol II remains largely unknown. Alternatively, the premature termination model proposes that Integrator removes promoter-proximal Pol II, leading to attenuation of gene expression.²²⁸ This model is supported by the finding that INTS11 depletion or mutation in fly cells results in gene upregulation.^228,233 Consistently, the rapid depletion of INTS11 in mouse embryonic stem cells also causes gene upregulation.²²⁹ Notably, the upregulated genes are enriched in short genes.²²⁹ It is likely that long genes are not upregulated upon INTS11 depletion due to a concomitant reduction in the elongation rate.²²⁹ However, how Integrator regulates elongation rates remains largely unknown. PAF1C and SPT6 have been shown to promote Integrator recruitment to genes,^164,173,221 suggesting a possible interplay between Integrator and elongation factors.

Although these two models appear to make opposing claims, there are some biological concepts in common. First, the endonuclease activity of Integrator is required for the removal of promoter-proximal Pol II in both models. Whether RNA cleavage by Integrator results in gene upregulation or downregulation may depend on the context of gene expression, such as rapid induction. Consistent with this idea, a recent study suggested that whether INTS11 loss promotes or attenuates productive elongation depends on the steady-state expression levels.²³¹ It remains unclear how RNA cleavage by Integrator could lead to Pol II removal. Perhaps upon cleavage an exposed 5’ end of nascent RNA is captured by the exonuclease XRN2, and the subsequent RNA degradation leads to Pol II eviction (Please refer to the review concerning this “torpedo” model ⁹¹). Second, Integrator is required for productive elongation in both models, at least in long genes. As we mentioned above, Integrator is functionally associated with various elongation factors including SEC, PAF1C and SPT6. In addition, Integrator’s protein phosphatase module may play a key role in the elongation process (we will discuss this below).

Protein phosphatase module of Integrator –

The recent identification of Integrator’s protein phosphatase activity^132,234,235 has dramatically expanded the scope of Integrator functions. The scaffold (A) and catalytic (C) subunits of type 2A protein phosphatase (PP2A) associate together with the Integrator subunits INTS6 and INTS8 to form the protein phosphatase module of Integrator.^132,234,235 Depletion of INTS6 or INTS8 results in dissociation of the PP2A-A and PP2A-C subunits from the Integrator complex and actively transcribed genes,^132,234,235 suggesting that INTS6/8 recruit PP2A-A and PP2A-C to transcribing Pol II. Several lines of evidence have revealed that the phosphatase and RNA endonuclease modules of Integrator are functionally distinct. Depletion of INTS6 or INTS8 does not affect the 3’-end processing of U snRNA transcripts,^234,235 suggesting that the RNA endonuclease module is still active without the phosphatase module. Consistently, the endonuclease module and scaffold subunits of Integrator can form a fully assembled complex without INTS6 or INTS8 subunits.²³⁵ Notably, the phosphatase module retains its activity without INTS11, the catalytic subunit of the RNA endonuclease module.²²⁹ Thus, the endonuclease and phosphatase modules of Integrator can be independently regulated during transcription.

The substrates of Integrator-PP2A include the Pol II CTD (Ser2, Thr4, Ser5, Ser7), SPT5 (Ser666, Thr806), and other Pol II-interacting factors.^132,234,235 Notably, phosphatase module substrates are also phosphorylated by CDK9, suggesting that Integrator-PP2A may counteract CDK9, and that small molecule activators of PP2A (SMAPs) could have the benefit of sensitizing cancers to CDK9 inhibition. Indeed, double treatment with the CDK9 inhibitor AZD’4573 and the PP2A activator DBK-1154 in a mouse xenograft model has shown synergistic anti-cancer activity in vivo.²³⁵

Consistent with the fact that Integrator-PP2A counteracts CDK9-dependent phosphorylation, Integrator-PP2A negatively regulates Pol II elongation. Depletion of INTS6 or INTS8 results in an increase in nascent mRNA and eRNA transcripts.^132,235 This is concomitant to an increase in elongation-associated phosphorylation of CTD Ser2 in both promoters and gene bodies.^132,235 INTS8 mutation that disrupts its association with PP2A results in the release of promoter-proximal paused Pol II,¹³² suggesting that PP2A plays a role in blocking Pol II release at 5’ end of genes. Whether Integrator-PP2A regulates transcription at the 3’ end of genes remains largely unknown.

BIOLOGICAL FUNCTIONS OF TRANSCRIPTION ELONGATION FACTORS

In this section, we discuss the biological functions of Pol II elongation factors and their role in pathogenesis of disease including developmental disorders, neurodegeneration, cancer, and aging. We highlight the functional link to these diseases beyond the detailed mechanisms of transcriptional elongation that we have discussed in the previous sections.

Function of CDK9-containing complexes in cancers and development–

In MLL-rearranged leukemia, studies have demonstrated that the protein fusion of MLL with an SEC subunit recruits SEC to MLL-fusion target genes, where SEC is required for aberrant transcriptional activation.^47,48 Rapid depletion of the SEC subunit ENL causes a loss of this aberrant SEC recruitment, attenuating leukemic growth.²³⁶ It should be noted that the anti-cancer effect of ENL depletion may be due to disruption of either or both of the two distinct ENL-containing complexes, SEC and DotCom (DOT1L Complex).^47,48,237 SEC is essential for transcription of the oncogene MYC and related genes.^51,61 The SEC scaffold subunit AFF4 was also found to be required for cell proliferation in pediatric brain cancer driven by the histone H3 lysine 27-methionine (H3K27M) mutation.^238,239 Together, these results led to the notion that the gene-specific function of SEC could be leveraged to downregulate oncogenic expression by targeted SEC inhibition, rather than by BRD4-CDK9 or total CDK9 inhibition. Disruption of the AFF4-Cyclin T interaction using the peptidomimetic inhibitors KL1/2 resulted in growth inhibition in both MYC-dependent cancer and H3K27M-dependent brain cancer.^61,238

SEC was also shown to be required for proper development in mammals. A chemical mutagenesis study in mice identified that a gain-of-function mutation in the N-terminal ALF domain of AFF1 causes neurodegeneration of the cerebellum.²⁴⁰ In humans, gain-of-function mutations in the N-terminal ALF domain of AFF4 have been linked to a developmental disorder involving cognitive impairment called CHOPS syndrome.²⁴¹ Mutations in the N-terminal ALF domain of AFF3 were recently shown to be associated with another type of human developmental disorder, KINSSHIP syndrome.²⁴² Notably, these ALF domain mutations of mouse AFF1 and human AFF3 or AFF4 all block proteasomal degradation mediated by the E3 ligase SIAH1, resulting in stabilization of the mutant proteins.^241-243 Thus, tight regulation of the AFF1/3/4 SEC scaffold protein stability is essential during development and for normal brain function. Downregulation of AFF2 (also known as FMR2) through hypermethylated triplet nucleotide expansions was also shown to be associated with intellectual disability,²⁴⁴ suggesting that dysregulation of SEC scaffold proteins plays a role in distinct developmental disorder etiologies. A recent study showed that the AFF1-containing SEC suppresses rapid response genes including ATF3 in undifferentiated human keratinocyte cells.²⁴⁵ This result suggests an additional role for SEC in epithelial progenitor differentiation that involves reprogramming of SEC function in response to developmental cues.

In addition to the functions of the BRD4-CDK9 complex in cancers that we discussed earlier, BRD4 has also been shown to play an important role in neural differentiation. A human genetic analysis revealed that loss-of-function mutations of BRD4 are associated with the neurodevelopmental disorder called Cornelia de Lange syndrome (CdLS).²⁴⁶ In mice, BRD4 loss in neural crest cells results in a developmental defect that phenocopies loss of the cohesin loader protein NIPBL.²⁴⁷ BRD4 physically interacts with NIPBL, which facilitates the cohesion-mediated process of chromatin loop extrusion. Importantly, loss of the cohesin antagonist WAPL can rescue the developmental defect observed in BRD4-deleted cells. These results suggest that BRD4 promotes neural differentiation through regulation of genome folding.

Functions of CDK12 in cancers –

Recurrent somatic alterations of the CDK12 gene have been identified in many cancer types, including ovarian cancers.²⁴⁸ Loss-of-function mutations of CDK12 observed in ovarian cancers are associated with genomic instability and sensitization to agents that induce DNA damage.^68,249,250 In breast cancers, amplifications of CDK12 are associated with splicing defects and invasiveness of cancer cells.⁶⁶ These observations led to the notion that dysregulation of CDK12’s function in transcriptional elongation control is involved in cancer pathogenesis, via various mechanisms depending on cancer type. The covalent CDK12/13 inhibitor THZ531 induces apoptotic cell death in leukemia cells.⁶⁹ The use of THZ531 in combination with other agents which inhibit DNA damage response or RNA processing could potentially help attenuate the growth of cancers that are “addicted” to CDK12’s effects on DNA damage response and RNA processing.

NELF’s function in development and cancers–

Pol II undergoes widespread pausing at promoter-proximal regions in mouse and fly embryos,^251,252 mouse embryonic stem cells²⁵³ and fly embryo-derived cells.²⁵⁴ In fly embryos, high levels of promoter-proximal Pol II have been linked to the synchronous expression of developmental genes.^255,256 Numerous studies illustrate the importance of NELF to developmental gene activation. In flies, NELF-deficient embryos show decreased expression of developmental gene reporters,²⁵⁷ and NELF knockdown impairs developmental hormone-induced gene activation.²⁵⁸ In zebrafish, NELF-B knockout leads to an improper hematopoiesis.²⁵⁹ In mice, NELF-B knockout causes early embryonic lethality,²⁶⁰ an insensitivity to developmental FGF/ERK signaling⁸² and defects in pluripotent epiblast state transitions.⁸⁶ Targeted degradation of NELF-B in mouse embryos at the pre-implantation stage impairs zygotic gene activation.²⁶¹ NELF-A may also function as a maternal factor to initiate zygotic gene expression, as it was shown to activate expression of the key developmental regulator gene Dux to drive the 2-cell-like state in mouse embryonic stem cells.²⁶² Together, these findings indicate a critical role for NELF in developmental gene activation.

A recent study revealed a functional connection between NELF-E and the development of breast cancer.²⁶³ Knockout of NELF-E in aggressive breast cancer cells results in selective downregulation of genes that promote tumorigenesis.²⁶³ NELF-E regulates expression of these genes through interaction with the transcription factor SLUG and transcriptional activation of the acetyltransferase KAT2B.²⁶³ These results suggest that breast cancer development may involve an alteration of gene expression that is mediated by NELF-E, SLUG and KAT2B, and that targeting this pathway is a potential therapeutic strategy for breast cancer.

Functions of SPT5/SPT4 in cancers, aging and neuronal development–

The oncogenic transcription factor MYC has been shown to regulate the interaction between SPT5 and Pol II.^114,264 MYC directly interacts with the N-terminal region of SPT5, which contains the conserved NGN domain.¹¹⁴ This MYC-SPT5 interaction can be disrupted by the activity of the CDK7 kinase, which assists handoff of SPT5 to Pol II.¹¹⁴ MYC may therefore increase the rate of elongation by facilitating CDK7-mediated SPT5 loading onto Pol II.¹¹⁴ However, the high levels of MYC which are frequently observed in cancer suppress SPT5 loading onto Pol II through the MYC-mediated sequestration of SPT5.¹¹⁴ Furthermore, multimerization of MYC under stress conditions sequesters SPT5 away from actively transcribing Pol II.²⁶⁴ These results led to the notion that the sequestration of SPT5 by MYC may be a general mechanism underlying the transcriptional repression associated with aggressive cancers. Intriguingly, a cancer subtype that possesses SUPTH5 (SPT5) gene amplification does not exhibit MYC-mediated transcriptional repression,¹¹⁴ suggesting that an excess supply of SPT5 can overcome sequestration by MYC and restore gene expression.

SPT4 is required for elongation of the long trinucleotide (CAG) repeat-containing transcripts that encode polyglutamine (polyQ) stretches, which are a characteristic feature of proteins associated with Huntington’s Disease (HD) and other neurological disorders, in yeast and human cells.²⁶⁵ Knockdown of SPT4 reduced polyQ reporter protein expression and expression of the HTT protein, leading to an overall decrease in aggregation of polyQ-containing proteins.²⁶⁵ Knockdown of SPT4 in the brain results in a reduction of Htt expression, prolonged lifespan and delay of motor impairment in a mouse HD model.²⁶⁶ SPT4 is also required for transcription of the C9orf72 hexanucleotide (GGGGCC) repeats associated with amyotrophic lateral sclerosis (ALS) and frontotemporal dementia.²⁶⁷ Knockdown of SPT4 in patient cells reduced expression of mRNA that contain this hexanucleotide repeat sequence and their resultant proteins.²⁶⁷ These observations indicate SPT4 inhibition as a potential therapeutic approach to suppress the expression of proteins that contain toxic repeat sequences. Because rapid depletion of SPT4 does not cause Pol II degradation,¹⁰⁸ SPT4 inhibition has the potential to selectively suppress transcription of pathogenic repeat sequences without affecting elongation at other genes. It will be necessary to develop potent compounds or other new SPT4-targeting approaches, both to suppress SPT4 function in patients and to advance the detailed mechanistic understanding of how SPT4 regulates the elongation of transcripts containing toxic repeat sequences.

SPT4 has been shown to promote aging in S. Cerevisiae by controlling the length of ribosomal DNA (rDNA) repeats, which shortens as cells age.²⁶⁸ Cells with SPT4 gene deletion exhibit stabilized rDNA repeats and decreased levels of non-coding RNA transcription from a promoter called E-pro within intergenic rDNA spacers.²⁶⁸ Although this non-coding RNA is nearly entirely silenced in young cells, its transcription is strongly activated in older cells in an SPT4-dependent manner.²⁶⁸ These observations suggest that non-coding RNA expression driven by SPT4 in older cells contributes to rDNA instability, thereby promoting cellular senescence, an irreversible state where cells can no longer divide.

The zebrafish foggy (SPT5 homolog) mutant is lethal due to its disrupted neuronal development, which includes the failure to develop distinct dopaminergic and other neuronal classes.¹⁰⁹ The mutation responsible for the foggy phenotype is the substitution of Foggy/Spt5 valine 1012, which is within the C-terminal domain but outside of the C-terminal repeat region (CTR), with aspartic acid (V1012D).¹⁰⁹ In vitro, the Foggy/Spt5 V1012D mutant protein results in an increased release of paused Pol II when CTR-phosphorylating CDK9 is inactivated, while Foggy functions to stimulate the rate of elongation as efficiently as wild-type when CDK9 is active.¹⁰⁹ This phenotype resembles the induced release of promoter-proximal Pol II observed in human cells with SPT5 CTR-TE (phospho-mimic) mutations,¹²⁵ implicating a functional interaction between the V1012D mutation and CTR hyperphosphorylation. These observations suggest that Spt5-mediated maintenance of the pausing state is required for development of distinct neuronal classes in zebrafish. However, which genes are directly regulated by Spt5 during neuronal development has remained unclear. Now that new techniques including rapid depletion and single-cell transcriptomic analysis are available, future studies may be able to determine how SPT5 regulates neuronal development in multicellular organisms.

The role of PAF1C in development and disease-

PAF1C has been shown to regulate developmental processes. For instance, the sunrise zebrafish mutant defective in the PAF1C subunit cdc73 gene was shown to exhibit severe developmental defects and embryonic lethality.¹⁶⁷ A null mutant of the PAF1C subunit gene rtf1 exhibited a similar morphological phenotype and embryonic lethality^167,269 and a null mutant of the PAF1C subunit leo1 showed defects in cardiac differentiation in zebrafish.²⁷⁰ In Drosophila, a null mutation of the PAF1C subunit Ctr9 was shown to cause defects in nervous system development and embryonic lethality.²⁷¹

PAF1C is also associated with disease states including cancer. For example, overexpression of the PAF1 protein promotes tumor growth.¹³⁸ Inactivating mutations of the tumor-suppressor protein CDC73 have been observed in a subset of cancer types including parathyroid cancers.¹³⁹ Although the L64P CDC73 mutant found in these cancers can interact with PAF1 in the manner of wild-type CDC73,²⁷² it fails to inhibit cyclin D1 expression in both NIH3T3 and HEK-293 cells.¹³⁹ A recent study revealed that PAF1C is critical to maintaining the survival of pancreatic cancers.²⁷³ PAF1 depletion results in accumulation of non-coding RNAs, including eRNAs and PROMPTs, for which transcription is driven by oncogenic Kras mutations.²⁷³ The accumulation of non-coding RNAs can cause DNA damage and cell death. These results suggest that pancreatic cancers are “addicted” to PAF1C-mediated suppression of non-coding RNA production. Recent advances in our understanding of how PAF1C functions in cancer development and stem cell maintenance have been summarized elsewhere.²⁷⁴

For infectious disease, PAF1C was shown to play a key role in antiviral response. The influenza viral protein NS1 directly interacts with PAF1C through its histone-mimic sequence.²⁷⁵ Through this interaction, NS1 hijacks Pol II elongation of the host cells, resulting in downregulation of antiviral gene expression.²⁷⁵ These results suggest that PAF1C activates antiviral gene expression, whereas NS1 suppresses this antiviral response through direct inhibition of PAF1C.

A recent study developed a first-in-class inhibitor of PAF1C (iPAF1C) and demonstrated that PAF1 inhibition could be used as a therapeutic approach to enable clearance of immune cells with latent human immunodeficiency virus-1 (HIV-1) infection.²⁷⁶ The general “shock and kill” strategy for HIV-1 treatment aims to trigger reactivation of transcriptionally silenced viral genes in HIV-1-infected immune cells, thereby achieving viral clearance in HIV-1 patients.²⁷⁷ iPAF1C treatment impedes PAF1C recruitment through disruption of the PAF1-CTR9 interaction and stimulates Pol II release, potentially triggering viral gene reactivation.²⁷⁶ Importantly, treatment of HIV-1-patient cells with iPAF1C enhances reactivation of latent proviruses in ex vivo HIV-1 patient cells.²⁷⁶

SPT6’s functions in development and cancers –

SPT6 mutant fly embryos have been shown to exhibit developmental arrests that cause lethality, presumably due to dysregulation of gene expression.^182,278 In humans, a recent study revealed that SPT6 functions to promote adult skin differentiation.²⁷⁹ Depletion of SPT6 results in loss of differentiation and conversion of epidermal cells to an intestinal-like state, primarily due to loss of expression of the epidermal fate regulator P63.²⁷⁹ These results suggest that SPT6-mediated elongation control is required for cell fate decisions during adult skin differentiation. In mouse embryonic stem cells, SPT6 was shown to maintain pluripotency via transcriptional regulation.²⁸⁰ Depletion of SPT6 results in selective downregulation of pluripotency factors including Nanog, with concomitant inactivation of super-enhancers associated with these pluripotency factor genes.²⁸⁰ Mechanistically, SPT6 counteracts Polycomb repressive complex 2 (PRC2) at super-enhancers through physical interaction with the PRC2 subunit Suz12.²⁸⁰

In brain cancers, SPT6 was found to be critical for self-renewal of glioblastoma cancer-stem like cells (GSCs).²⁸¹ Depletion of SPT6 in GSCs results in loss of the capacity to initiate cancers, primarily due to decreased expression of DNA damage response genes and greater genomic instability.²⁸¹ These results suggest that SPT6 functions to support survival of aggressive brain cancers through upregulation of DNA damage response pathways that maintain genomic stability.

Functions of Integrator in development and genomic stability –

It has been shown that Integrator is required for the generation of cilia.^282-284 A recent human genetic study identified germline loss-of-function mutations of INTS13 in Oral-facial-digital (OFD) syndromes, a subtype of developmental ciliopathies.²⁸⁵ These INTS13 mutations disrupt Integrator complex assembly.²⁸⁵ Depletion of INTS13 results in ciliogenesis defects in human cells and Xenopus embryos, suggesting a disease mechanism that involves INTS13 mutations.²⁸⁵ INTS13 was also shown to be required for human myeloid differentiation and to function via enhancer activation.²⁸⁶ While INTS13 is not required for the endonuclease activity of Integrator, INTS13 does act as a mediator of cell lineage-determining transcription factors.²⁸⁶ Another genetic study identified INTS1 and INTS8 mutations in human neurodevelopmental syndromes,²⁸⁷ suggesting that Integrator functions in human brain development.

Integrator also plays a key role in genomic stability. A recent study revealed that Integrator physically interacts with the DNA damage sensor SOSS via the INTS3 subunit, and that the SOSS subunit SSB1 forms condensates with Integrator.²⁸⁸ Integrator loss results in increased formation of the DNA-RNA duplex R-loop structures that can induce genomic instability.²⁸⁸ Thus, Integrator functions to protect cells from genomic instability through interaction with the DNA damage sensor and formation of nuclear condensates.

Mechanism of aging: alteration of transcriptional elongation properties over a lifetime–

The process of aging, a gradual loss of biological function throughout lifespan, has been linked to changes in the epigenetic landscape including altered levels and repositioning of DNA methylation and histone modifications.²⁸⁹ Until recently, however, it was largely unclear whether the aging process might also involve alteration of Pol II elongation. We briefly summarize recent findings regarding the relationship between transcriptional elongation and aging and provide some future perspective for studies looking at the aging process and the role of the elongation factors described above.

A role for Pol II elongation rate in the mechanism of aging–

A recent study provided a catalog of Pol II elongation rates in both young and old model organisms.²⁹⁰ High-throughput rate measurements were achieved simply using total RNA-seq data,²⁹⁰ based on the previous finding that intronic RNA-seq signal exhibits a saw-tooth pattern, the slope of which can be used to compute relative changes in Pol II elongation rate (Figure 6A).^21,291 This study demonstrated that rates of Pol II elongation increase upon aging in human, rat, mouse, fly and worm.²⁹⁰ Notably, flies and worms with “slow Pol II” due to mutations in a Pol II subunit exhibited an extended lifespan.²⁹⁰ This genetic evidence strongly suggests that an elevated Pol II elongation rate is not only a conserved feature of aging but also part of a conserved aging mechanism. It also suggests that lifespan could be extended by slowing down Pol II in other species (Figure 6B). It remains unclear how a “fast Pol II” could cause or worsen the effects of aging, but increasing Pol II speed over lifespan is attributed in part to a partial loss of histone. Deprivation of nucleosome occupancy has been observed in aged organisms from yeast to human.^292,293 In yeast, histone loss upon aging resulted in gene upregulation²⁹⁴ and overexpression of histone proteins extended lifespan.²⁹² Consistent with these precedent studies, overexpression of histone H3 was shown to reduce Pol II elongation rates and suppress senescence in human cells and to extend lifespan in flies.²⁹⁰ These results suggest an aging model in which 1) a decreasing supply of histone proteins causes partial loss of nucleosome occupancy, 2) rates of Pol II elongation are increased as a result of loss of nucleosomal obstacles to transcription in gene bodies, and 3) increased Pol II speed causes or worsens the effects of aging through unknown mechanisms. Given that Pol II elongation rate is regulated by various elongation factors including SPT5, PAF1C, SPT6 and SEC, development and use of potent inhibitors targeting elongation factors will be a new direction of future studies aiming to address the challenge of slowing down aging processes. Increasing histone gene expression may also be an alternative way to slow down Pol II to “slow down aging”.

Figure 6. Alteration of Pol II elongation over lifetime.

(A) Schematic of a saw-tooth pattern in intronic RNA-seq signal. (B) Changes in Pol II elongation rates and the abundance of long transcripts over lifetime are indicated. Slowing down the Pol II elongation rate results in extended lifespan.

Loss of long transcripts as you age –

On the other hand, a strong correlation between aging and lower abundance of “long transcripts” has recently been demonstrated (Figure 6B).^295-297 In an accelerated-aging model, mice which possess mutations in the nucleotide excision DNA repair pathway exhibit a preferential loss of mRNA expression from long genes in liver tissues, while dietary restrictions that increase their lifespan rescue this length-associated downregulation.²⁹⁶ Consistently, the aged mouse liver was shown to exhibit an approximately 1.5-fold reduction in total RNA synthesis.²⁹⁷ Productive elongation declined and Ser2P-associated Pol II occupancy also increased in a gene length-dependent manner in the aged mouse liver,²⁹⁷ suggesting that Pol II becomes stalled in long genes upon aging. The mechanism of this length-associated Pol II stalling can be explained by the stochasticity of DNA damage events in the genome: long genes statistically accumulate more DNA lesions, which pose obstacles to productive transcription, than short genes purely as a function of their greater length. Consistent with this model, the study showed that UV-induced DNA damage caused transcriptional stalling in long genes, and that the normally aged liver exhibited an activation of DNA repair pathway.²⁹⁷ These results strongly suggest that aging is associated with Pol II stalling in long genes, but it is currently unknown whether this association is causal, and if so whether Pol II stalling in long genes is a cause or a consequence of aging. Therapeutic approaches to aging can vary depending on these conclusions. Although Pol II elongation speed increases upon aging, it is unclear whether Pol II stalling in long genes and a fast Pol II can co-exist in aged animal. As one possible complication, slowing down Pol II elongation rates to extend lifespan might affect the efficiency of the DNA damage repair mechanisms that are required for productive transcription in long genes, with increasing requirement as lifespan extends.

Contributions of Pol II elongation factors to aging-related phenotypes –

The loss of long transcripts which is a hallmark of aged cells is reminiscent of the processivity defects observed in PAF1C, SPT6, or CDK12-deficient cells (please see the SPT6 section). It is tempting to speculate that regulation of processivity by PAF1C, SPT6 or CDK12/13 might also be impaired upon aging. In this scenario, loss of long transcripts might be attributed not only to DNA lesions accumulated over time but also to processivity defects due to dysregulation of Pol II elongation factors. It should be noted that these two causal processes are not mutually exclusive, as CDK12 inhibition results in the downregulation of DNA damage response genes and sensitization to DNA damage agents (Please see the CDK9 section). Alternatively, Pol II stalling in aged cells could also be explained by a loss of factors that rescue stalled Pol II, such as TFIIS. The combination of spt6 mutation and TFIIS gene deletion results in synthetic lethality in yeast cells, and SPT6 recruits TFIIS through interaction with IWS1 in human cells (please see the SPT6 section). These observations suggest a conserved mechanism that might prevent Pol II stalling through the cooperative function of SPT6 and TFIIS. If the functions of SPT6, IWS1 or TFIIS became impaired over the course of a lifetime, such impairment could lead to Pol II stalling similar to that observed in aged cells. Going forward, it will be necessary to characterize the functions of Pol II elongation factors in aged versus young cells and to assess the potential contributions of elongation factors to aging-related transcriptional alterations.

CONCLUDING REMARKS

We published in 1996 the first link between transcriptional elongation control and human cancer,¹ demonstrating that the elongation stage of transcription is a key regulatory step in development and during disease pathogenesis. Since then, we have demonstrated that diverse translocation within the MLL gene causing leukemia is into many subunits of the Super Elongation Complex, regulating transcriptional elongation control and pause release and its misregulation causing cancer.^2,3 Collectively, our studies and that of colleagues in the field as described in this review have demonstrated the importance of transcription elongation control not only in developmental regulation of gene expression, but also during disease pathogenesis and in aging. In this review, we present a current understanding of the molecular mechanisms involved in Pol II elongation control. These mechanisms include recruitment of elongation factors to Pol II, swapping of elongation factors, allosteric regulation of Pol II processivity and Pol II stabilization. We have discussed the biochemical and mechanistic details of elongation factor function that have recently emerged thanks to the development of advanced techniques such as rapid depletion, nascent RNA sequencing and cryo-EM analysis of protein complex structure. We have also attempted to reconcile several discrepancies between recent in vivo findings and earlier in vitro biochemical evidence regarding promoter-proximal pause release and elongation processivity. Finally, we have highlighted the functional link between dysregulation of transcriptional elongation and various human diseases.

Dysregulated transcriptional elongation has long been associated both with developmental defects and with pathological processes including cancer and infectious disease. Now, accumulating evidence suggests for a critical involvement of dysregulated Pol II elongation in the aging process, underscoring the importance of elongation control. Notably, it has been demonstrated that slowing down the Pol II elongation rate results in extended lifespan in model systems. However, it is entirely unknown how changes to the Pol II rate might affect macro aging processes, whether the function of elongation factors might change over the course of a lifetime, or which specific tissues might play a role in the macro aging processes regulated by Pol II elongation rates. Although aging is also associated with the loss of long transcripts, the existence or direction of the causal relationship between the two remains unclear. Nonetheless, achieving precise control of Pol II dynamics will be crucial to the development of transcriptional approaches to therapeutic extend lifespan extension. Precise control would require modulation of Pol II rate while also maintaining processivity in difficult sequence environments to avoid significant loss of long transcripts. Future studies that solidify a fundamental understanding of transcriptional elongation regulation are therefore the key to successful development not only of anticancer drugs, but also of safe and feasible therapies to potentially slow or ease the aging process.